{ "api_version": "1", "updated": "2026-11-10", "latest_event": "2026-11-09", "rights": "© Frontier Log, all rights reserved.", "benchmark": { "slug": "swe-bench-pro", "name": "SWE-bench Pro", "family": "swe-bench", "unit": "%", "higher_is_better": true, "status": "active", "active": true, "superseded_by": null, "retired": null, "benchmark_url": "https://scale.com/leaderboard/swe_bench_pro_public", "introduced": "2025-09-19", "creator": "Scale AI", "introduced_source": "https://scale.com/blog/swe-bench-pro", "settings": [ { "setting": "public", "name": "Public set (731 tasks)", "url": "https://frontierlog.ai/benchmark/swe-bench-pro", "results": 29 }, { "setting": "corrected", "name": "Lab's corrected public set", "url": "https://frontierlog.ai/benchmark/swe-bench-pro/corrected", "results": 10 }, { "setting": "unstated", "name": "Set not stated", "url": "https://frontierlog.ai/benchmark/swe-bench-pro/unstated", "results": 46 } ] }, "total": 46, "results": [ { "id": "swe-1-5:swe-bench-pro:unstated:1", "model": "swe-1-5", "model_name": "SWE-1.5", "model_date": "2025-10-29", "event": "https://frontierlog.ai/event/2025-10-29-swe-1-5", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 34.47, "unit": "%", "printed": "34.47%", "reported_by": "cognition", "run_by": null, "label": "SWE-agent", "source": "https://cognition.com/blog/swe-1-5", "source_title": "Introducing SWE-1.5: Our Fast Agent Model", "source_date": "2025-10-29", "where": "'The Agent-Model Interface' section, bar chart 'SWE-Bench Pro: SWE-1.5 on Different Harnesses', value label above the bar 'SWE-agent'", "own": true, "primary": true, "parallel": false, "notes": "SWE-agent harness", "caveat": "", "human_verified": true, "human_verified_on": "2026-11-07", "check": "human", "corroborated": false, "corroborated_on": null, "corroborated_by": [], "source_image": "https://frontierlog.ai/sources/35a0bdb52c51997d.png", "source_image_original": "https://cdn.sanity.io/images/2mc9cv2v/production/065263aa9933f9196d9fca336d42a975d199c94e-2550x1500.png?w=1600&fit=max" }, { "id": "swe-1-5:swe-bench-pro:unstated:2", "model": "swe-1-5", "model_name": "SWE-1.5", "model_date": "2025-10-29", "event": "https://frontierlog.ai/event/2025-10-29-swe-1-5", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 29.0, "unit": "%", "printed": "29.00%", "reported_by": "cognition", "run_by": null, "label": "Claude Code", "source": "https://cognition.com/blog/swe-1-5", "source_title": "Introducing SWE-1.5: Our Fast Agent Model", "source_date": "2025-10-29", "where": "'The Agent-Model Interface' section, bar chart 'SWE-Bench Pro: SWE-1.5 on Different Harnesses', value label above the bar 'Claude Code'", "own": true, "primary": false, "parallel": false, "notes": "Claude Code harness", "caveat": "", "human_verified": true, "human_verified_on": "2026-11-07", "check": "human", "corroborated": false, "corroborated_on": null, "corroborated_by": [], "source_image": "https://frontierlog.ai/sources/35a0bdb52c51997d.png", "source_image_original": "https://cdn.sanity.io/images/2mc9cv2v/production/065263aa9933f9196d9fca336d42a975d199c94e-2550x1500.png?w=1600&fit=max" }, { "id": "claude-opus-4-5:swe-bench-pro:unstated:1", "model": "claude-opus-4-5", "model_name": "Claude Opus 4.5", "model_date": "2025-11-24", "event": "https://frontierlog.ai/event/2025-11-24-claude-opus-4-5", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 52.0, "unit": "%", "printed": "52.0%", "reported_by": "anthropic", "run_by": null, "label": "Claude Opus 4.5 (no thinking)", "source": "https://www.anthropic.com/claude-opus-4-5-system-card", "source_title": "Claude Opus 4.5 System Card", "source_date": "2025-11-24", "where": "Section 2.4, Table 2.4.A, column 'SWE-bench Pro', row 'Claude Opus 4.5 (no thinking)'", "own": true, "primary": true, "parallel": false, "notes": "no thinking; average of 5 trials", "caveat": "", "human_verified": true, "human_verified_on": "2026-10-19", "check": "human", "corroborated": false, "corroborated_on": null, "corroborated_by": [], "source_capture": "https://frontierlog.ai/captures/ceb9c141f42aab5e.png" }, { "id": "claude-opus-4-5:swe-bench-pro:unstated:2", "model": "claude-opus-4-5", "model_name": "Claude Opus 4.5", "model_date": "2025-11-24", "event": "https://frontierlog.ai/event/2025-11-24-claude-opus-4-5", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 51.6, "unit": "%", "printed": "51.60%", "reported_by": "anthropic", "run_by": null, "label": "Claude Opus 4.5 (64k thinking)", "source": "https://www.anthropic.com/claude-opus-4-5-system-card", "source_title": "Claude Opus 4.5 System Card", "source_date": "2025-11-24", "where": "Section 2.4, Table 2.4.A, column 'SWE-bench Pro', row 'Claude Opus 4.5 (64k thinking)'", "own": true, "primary": false, "parallel": false, "notes": "64K thinking budget; average of 5 trials", "caveat": "", "human_verified": true, "human_verified_on": "2026-10-19", "check": "human", "corroborated": false, "corroborated_on": null, "corroborated_by": [], "source_capture": "https://frontierlog.ai/captures/ceb9c141f42aab5e.png" }, { "id": "minimax-m2-1:swe-bench-pro:unstated:1", "model": "minimax-m2-1", "model_name": "MiniMax M2.1", "model_date": "2025-12-23", "event": "https://frontierlog.ai/event/2025-12-23-minimax-m2-1", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 49.7, "unit": "%", "printed": "49.7", "reported_by": "minimax", "run_by": null, "label": "MiniMax M2.1", "source": "https://www.minimax.io/news/minimax-m25", "source_title": "MiniMax M2.5: Built for Real-World Productivity.", "source_date": "2026-02-12", "where": "later report: MiniMax M2.5 post, first benchmark chart image (eight panels), panel 'SWE-Bench Pro', second bar (legend 'MiniMax M2.1'), value label", "own": false, "primary": true, "parallel": false, "notes": "Claude Code harness; average of 4 runs", "caveat": "", "human_verified": true, "human_verified_on": "2026-10-20", "check": "human", "corroborated": false, "corroborated_on": null, "corroborated_by": [], "source_image": "https://frontierlog.ai/sources/7eaf807b2996051e.png", "source_image_original": "https://file.cdn.minimax.io/public/97f76950-2c60-4a9b-bb96-228454afabe9.png" }, { "id": "kimi-k2-5:swe-bench-pro:unstated:1", "model": "kimi-k2-5", "model_name": "Kimi K2.5", "model_date": "2026-01-27", "event": "https://frontierlog.ai/event/2026-01-27-kimi-k2-5", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 50.7, "unit": "%", "printed": "50.7", "reported_by": "moonshot", "run_by": null, "label": "Kimi K2.5 (Thinking)", "source": "https://www.kimi.ai/blog/kimi-k2-5", "source_title": "Kimi K2.5: Visual Agentic Intelligence", "source_date": "2026-01-27", "where": "Appendix 'Benchmark table', row 'SWE-Bench Pro' (Coding), column 'Kimi K2.5 (Thinking)' (read from the page's embedded benchmark JSON and the rendered table)", "own": true, "primary": true, "parallel": false, "notes": "in-house minimal-tool harness; non-thinking mode; average of 5 runs", "caveat": "", "human_verified": true, "human_verified_on": "2026-10-17", "check": "human", "corroborated": true, "corroborated_on": "2026-10-19", "corroborated_by": [ { "url": "https://www.kimi.ai/blog/kimi-k2-6", "name": "Kimi K2.6", "title": "Kimi K2.6: Advancing Open-Source Coding", "lab": "moonshot", "date": "2026-04-20", "printed": "50.7", "where": "Benchmark Table (appendix), row SWE-Bench Pro ; column Kimi K2.5" }, { "url": "https://thinkingmachines.ai/news/introducing-inkling/", "name": "Inkling", "title": "Inkling: Our Open-Weights Model", "lab": "thinking-machines", "date": "2026-07-15", "printed": "50.7%", "where": "'Benchmarking Inkling' main table, row 'SWEBench Pro Public', column 'Kimi K2.5'" } ], "source_capture": "https://frontierlog.ai/captures/a8cc819ddd0da74b.png" }, { "id": "qwen3-coder-next:swe-bench-pro:unstated:1", "model": "qwen3-coder-next", "model_name": "Qwen3-Coder-Next", "model_date": "2026-02-03", "event": "https://frontierlog.ai/event/2026-02-03-qwen3-coder-next", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 44.3, "unit": "%", "printed": "44.3", "reported_by": "alibaba", "run_by": null, "label": "Qwen3-Coder-Next", "source": "https://qwen.ai/blog?id=qwen3-coder-next", "source_title": "Qwen3-Coder-Next: Pushing Small Hybrid Models on Agentic Coding", "source_date": "2026-02-03", "where": "'Performance on Coding Agent Benchmarks' chart image (benchmarks.png), group 'SWE-Bench Pro (w/ SWE-Agent)', first bar (legend 'Qwen3-Coder-Next (Ours)'), value label", "own": true, "primary": true, "parallel": false, "notes": "SWE-Agent harness", "caveat": "", "human_verified": true, "human_verified_on": "2026-10-20", "check": "human", "corroborated": false, "corroborated_on": null, "corroborated_by": [], "source_image": "https://frontierlog.ai/sources/2be7fdf646f0b0e5.png", "source_image_original": "https://qianwen-res.oss-accelerate-overseas.aliyuncs.com/Qwen3-Coder-Next/benchmarks.png" }, { "id": "claude-opus-4-6:swe-bench-pro:unstated:1", "model": "claude-opus-4-6", "model_name": "Claude Opus 4.6", "model_date": "2026-02-05", "event": "https://frontierlog.ai/event/2026-02-05-claude-opus-4-6", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 53.4, "unit": "%", "printed": "53.4%", "reported_by": "anthropic", "run_by": null, "label": "Opus 4.6", "source": "https://www.anthropic.com/news/claude-opus-4-7", "source_title": "Introducing Claude Opus 4.7", "source_date": "2026-04-16", "where": "benchmark table (image), row 'Agentic coding, SWE-bench Pro', column 'Opus 4.6'", "own": false, "primary": true, "parallel": false, "notes": "memorization screens flag a subset of problems", "caveat": "", "human_verified": true, "human_verified_on": "2026-10-19", "check": "human", "corroborated": true, "corroborated_on": "2026-10-19", "corroborated_by": [ { "url": "https://www.anthropic.com/claude-opus-4-7-system-card", "name": "Claude Opus 4.7 card", "title": "Claude Opus 4.7 System Card", "lab": "anthropic", "date": "2026-04-16", "printed": "53.4%", "where": "section 8.1 Evaluation summary table, SWE-bench Pro row, Claude Opus 4.6 column" } ], "source_image": "https://frontierlog.ai/sources/527cd462f89b6de4.png", "source_image_original": "https://www-cdn.anthropic.com/images/4zrzovbb/website/d434d15757c6abac1122af483617741776d5a114-2600x2638.png" }, { "id": "glm-5:swe-bench-pro:unstated:1", "model": "glm-5", "model_name": "GLM-5", "model_date": "2026-02-12", "event": "https://frontierlog.ai/event/2026-02-12-glm-5", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 55.1, "unit": "%", "printed": "55.1", "reported_by": "zai", "run_by": null, "label": "GLM-5", "source": "https://z.ai/blog/glm-5.1", "source_title": "GLM-5.1: Towards Long-Horizon Tasks", "source_date": "2026-04-07", "where": "benchmark table near the end, row 'SWE-Bench Pro', column 'GLM-5'", "own": false, "primary": true, "parallel": false, "notes": "OpenHands with a tailored prompt", "caveat": "", "human_verified": true, "human_verified_on": "2026-10-19", "check": "human", "corroborated": true, "corroborated_on": "2026-10-19", "corroborated_by": [ { "url": "https://qwen.ai/blog?id=qwen3.6", "name": "Qwen3.6-Plus", "title": "Qwen3.6-Plus: Towards Real World Agents", "lab": "alibaba", "date": "2026-04-02", "printed": "55.1", "where": "Score chart image (qwen3.6_plus_score.png), panel SWE-bench Pro, value label on the GLM5 bar (bars identified by legend colour and logo)" } ], "source_capture": "https://frontierlog.ai/captures/35c66845b36400f8.png" }, { "id": "minimax-m2-5:swe-bench-pro:unstated:1", "model": "minimax-m2-5", "model_name": "MiniMax M2.5", "model_date": "2026-02-12", "event": "https://frontierlog.ai/event/2026-02-12-minimax-m2-5", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 55.4, "unit": "%", "printed": "55.4", "reported_by": "minimax", "run_by": null, "label": "MiniMax M2.5", "source": "https://www.minimax.io/news/minimax-m25", "source_title": "MiniMax M2.5: Built for Real-World Productivity.", "source_date": "2026-02-12", "where": "first benchmark chart image (top of the post, eight panels), panel 'SWE-Bench Pro', bar 'MiniMax M2.5' (leftmost), value label; the same label appears in the later coding chart image 54ddb070-….png", "own": true, "primary": true, "parallel": false, "notes": "Claude Code harness; average of 4 runs", "caveat": "", "human_verified": true, "human_verified_on": "2026-10-17", "check": "human", "corroborated": true, "corroborated_on": "2026-10-19", "corroborated_by": [ { "url": "https://www.minimax.io/news/minimax-m27-en", "name": "MiniMax M2.7", "title": "MiniMax M2.7: Early Echoes of Self-Evolution", "lab": "minimax", "date": "2026-03-18", "printed": "55.4", "where": "Bar chart image (8 panels); panel 'SWE Bench Pro'; value label on M2.5 bar" } ], "source_image": "https://frontierlog.ai/sources/7eaf807b2996051e.png", "source_image_original": "https://file.cdn.minimax.io/public/97f76950-2c60-4a9b-bb96-228454afabe9.png" }, { "id": "claude-sonnet-4-6:swe-bench-pro:unstated:1", "model": "claude-sonnet-4-6", "model_name": "Claude Sonnet 4.6", "model_date": "2026-02-17", "event": "https://frontierlog.ai/event/2026-02-17-claude-sonnet-4-6", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 58.1, "unit": "%", "printed": "58.1%", "reported_by": "anthropic", "run_by": null, "label": "Sonnet 4.6", "source": "https://www.anthropic.com/news/claude-sonnet-5", "source_title": "Introducing Claude Sonnet 5", "source_date": "2026-06-30", "where": "benchmark table (image), row 'Agentic coding, SWE-bench Pro', column 'Sonnet 4.6'", "own": false, "primary": true, "parallel": false, "notes": "", "caveat": "", "human_verified": true, "human_verified_on": "2026-10-19", "check": "human", "corroborated": true, "corroborated_on": "2026-10-19", "corroborated_by": [ { "url": "https://www.anthropic.com/claude-sonnet-5-system-card", "name": "Claude Sonnet 5 card", "title": "Claude Sonnet 5 System Card", "lab": "anthropic", "date": "2026-06-30", "printed": "58.1", "where": "Table 8.1.A, SWE-bench Pro row, Sonnet 4.6 column" } ], "source_image": "https://frontierlog.ai/sources/849d6a908f0d46ac.png", "source_image_original": "https://www-cdn.anthropic.com/images/4zrzovbb/website/9941d610909f28a504e16dd5af823df172ec6035-2600x1234.png" }, { "id": "minimax-m2-7:swe-bench-pro:unstated:1", "model": "minimax-m2-7", "model_name": "MiniMax M2.7", "model_date": "2026-03-18", "event": "https://frontierlog.ai/event/2026-03-18-minimax-m2-7", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 56.22, "unit": "%", "printed": "56.22%", "reported_by": "minimax", "run_by": null, "label": "M2.7", "source": "https://www.minimax.io/news/minimax-m27-en", "source_title": "MiniMax M2.7: Early Echoes of Self-Evolution", "source_date": "2026-03-18", "where": "opening list, item 1 ('On the SWE-Pro benchmark, M2.7 scored …'); repeated in 'Professional Software Engineering'", "own": true, "primary": true, "parallel": false, "notes": "", "caveat": "", "human_verified": true, "human_verified_on": "2026-10-19", "check": "human", "corroborated": true, "corroborated_on": "2026-10-19", "corroborated_by": [ { "url": "https://www.minimax.io/blog/minimax-m3", "name": "MiniMax M3", "title": "MiniMax M3: Frontier Coding, 1M Context, Native Multimodality — All in One Model", "lab": "minimax", "date": "2026-06-01", "printed": "56.2", "where": "Full benchmark table image; row 'SWE-Bench Pro'; column 'MiniMax M2.7'" }, { "url": "https://z.ai/blog/glm-5.1", "name": "GLM-5.1", "title": "GLM-5.1: Towards Long-Horizon Tasks", "lab": "zai", "date": "2026-04-07", "printed": "56.2", "where": "Benchmark table, row 'SWE-Bench Pro', column 'MiniMax M2.7'" } ] }, { "id": "minimax-m2-7:swe-bench-pro:unstated:2", "model": "minimax-m2-7", "model_name": "MiniMax M2.7", "model_date": "2026-03-18", "event": "https://frontierlog.ai/event/2026-03-18-minimax-m2-7", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 56.2, "unit": "%", "printed": "56.2", "reported_by": "minimax", "run_by": null, "label": "MiniMax M2.7", "source": "https://www.minimax.io/blog/minimax-m3", "source_title": "MiniMax M3: Frontier Coding, 1M Context, Native Multimodality — All in One Model", "source_date": "2026-06-01", "where": "MiniMax M3 post, benchmark table image (tall, nine columns), row 'SWE-Bench Pro', column 'MiniMax M2.7' (second column)", "own": false, "primary": false, "parallel": false, "notes": "", "caveat": "", "human_verified": true, "human_verified_on": "2026-10-19", "check": "human", "corroborated": true, "corroborated_on": "2026-10-19", "corroborated_by": [ { "url": "https://www.minimax.io/news/minimax-m27-en", "name": "MiniMax M2.7", "title": "MiniMax M2.7: Early Echoes of Self-Evolution", "lab": "minimax", "date": "2026-03-18", "printed": "56.22%", "where": "Professional Software Engineering paragraph: 'On SWE-Pro ... M2.7 achieved a 56.22% accuracy rate'" }, { "url": "https://z.ai/blog/glm-5.1", "name": "GLM-5.1", "title": "GLM-5.1: Towards Long-Horizon Tasks", "lab": "zai", "date": "2026-04-07", "printed": "56.2", "where": "Benchmark table, row 'SWE-Bench Pro', column 'MiniMax M2.7'" } ], "source_image": "https://frontierlog.ai/sources/5720f3ad06ea2ce9.jpg", "source_image_original": "https://filecdn.minimax.chat/public/img_v3_02128_b7726cd8-879a-4b7a-a9da-db4395ea597g-1780272508686.jpg" }, { "id": "claude-mythos-preview:swe-bench-pro:unstated:1", "model": "claude-mythos-preview", "model_name": "Claude Mythos Preview", "model_date": "2026-04-07", "event": "https://frontierlog.ai/event/2026-04-07-claude-mythos-preview", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 77.8, "unit": "%", "printed": "77.8%", "reported_by": "anthropic", "run_by": null, "label": "Mythos Preview", "source": "https://www.anthropic.com/news/claude-opus-4-7", "source_title": "Introducing Claude Opus 4.7", "source_date": "2026-04-16", "where": "benchmark table (image), row 'Agentic coding, SWE-bench Pro', column 'Mythos Preview'", "own": false, "primary": true, "parallel": false, "notes": "", "caveat": "", "human_verified": true, "human_verified_on": "2026-10-17", "check": "human", "corroborated": true, "corroborated_on": "2026-10-19", "corroborated_by": [ { "url": "https://www.anthropic.com/claude-fable-5-mythos-5-system-card", "name": "Claude Fable 5 and Mythos 5 card", "title": "Claude Fable 5 and Claude Mythos 5 System Card", "lab": "anthropic", "date": "2026-06-09", "printed": "77.8", "where": "Table 8.1.A (p.251), row SWE-bench Pro, column Mythos Preview" }, { "url": "https://openai.com/index/gpt-5-6", "name": "GPT-5.6", "title": "GPT-5.6: Frontier intelligence that scales with your ambition", "lab": "openai", "date": "2026-07-09", "printed": "77.8%", "where": "table 1 \"SWE-Bench Pro\" row, \"Claude Mythos Preview\" column" } ], "source_image": "https://frontierlog.ai/sources/527cd462f89b6de4.png", "source_image_original": "https://www-cdn.anthropic.com/images/4zrzovbb/website/d434d15757c6abac1122af483617741776d5a114-2600x2638.png" }, { "id": "glm-5-1:swe-bench-pro:unstated:1", "model": "glm-5-1", "model_name": "GLM-5.1", "model_date": "2026-04-07", "event": "https://frontierlog.ai/event/2026-04-07-glm-5-1", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 58.4, "unit": "%", "printed": "58.4", "reported_by": "zai", "run_by": null, "label": "GLM-5.1", "source": "https://z.ai/blog/glm-5.1", "source_title": "GLM-5.1: Towards Long-Horizon Tasks", "source_date": "2026-04-07", "where": "benchmark table near the end, row 'SWE-Bench Pro', column 'GLM-5.1'", "own": true, "primary": true, "parallel": false, "notes": "OpenHands with a tailored prompt; 200K context", "caveat": "", "human_verified": true, "human_verified_on": "2026-10-19", "check": "human", "corroborated": true, "corroborated_on": "2026-10-19", "corroborated_by": [ { "url": "https://api-docs.deepseek.com/news/news260424/", "name": "DeepSeek V4 Preview", "title": "DeepSeek V4 Preview Release", "lab": "deepseek", "date": "2026-04-24", "printed": "58.4", "where": "Full benchmark table image; row 'SWE Pro (Resolved)'; column 'GLM-5.1 Thinking'" }, { "url": "https://www.minimax.io/blog/minimax-m3", "name": "MiniMax M3", "title": "MiniMax M3: Frontier Coding, 1M Context, Native Multimodality — All in One Model", "lab": "minimax", "date": "2026-06-01", "printed": "58.4", "where": "Full benchmark table image; row 'SWE-Bench Pro'; column 'GLM 5.1 Thinking'" }, { "url": "https://qwen.ai/blog?id=qwen3.6-max-preview", "name": "Qwen3.6-Max-Preview", "title": "Qwen3.6-Max-Preview: Smarter, Sharper, Still Evolving", "lab": "alibaba", "date": "2026-04-18", "printed": "58.4", "where": "Score chart image (qwen3.6_max_preview_score.png), panel SWE-bench Pro, value label on the GLM 5.1 bar" }, { "url": "https://z.ai/blog/glm-5.2", "name": "GLM-5.2", "title": "GLM-5.2: Built for Long-Horizon Tasks", "lab": "zai", "date": "2026-06-16", "printed": "58.4", "where": "Prose: '62.1 vs. 58.4 on SWE-bench Pro' (GLM-5.1)" } ], "source_capture": "https://frontierlog.ai/captures/35c66845b36400f8.png" }, { "id": "claude-opus-4-7:swe-bench-pro:unstated:1", "model": "claude-opus-4-7", "model_name": "Claude Opus 4.7", "model_date": "2026-04-16", "event": "https://frontierlog.ai/event/2026-04-16-claude-opus-4-7", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 64.3, "unit": "%", "printed": "64.3%", "reported_by": "anthropic", "run_by": null, "label": "Opus 4.7", "source": "https://www.anthropic.com/news/claude-opus-4-7", "source_title": "Introducing Claude Opus 4.7", "source_date": "2026-04-16", "where": "benchmark table (image), row 'Agentic coding, SWE-bench Pro', column 'Opus 4.7'", "own": true, "primary": true, "parallel": false, "notes": "memorization screens flag a subset of problems", "caveat": "", "human_verified": true, "human_verified_on": "2026-10-17", "check": "human", "corroborated": true, "corroborated_on": "2026-10-19", "corroborated_by": [ { "url": "https://www.anthropic.com/claude-opus-4-7-system-card", "name": "Claude Opus 4.7 card", "title": "Claude Opus 4.7 System Card", "lab": "anthropic", "date": "2026-04-16", "printed": "64.3%", "where": "Table 8.1.A (p.191), row SWE-bench Pro, column Claude Opus 4.7" }, { "url": "https://www.anthropic.com/claude-opus-4-8-system-card", "name": "Claude Opus 4.8 card", "title": "Claude Opus 4.8 System Card", "lab": "anthropic", "date": "2026-05-28", "printed": "64.3", "where": "Table 8.1.A (p.194), row SWE-bench Pro, column Claude Opus 4.7" }, { "url": "https://www.anthropic.com/news/claude-opus-4-8", "name": "Claude Opus 4.8", "title": "Introducing Claude Opus 4.8", "lab": "anthropic", "date": "2026-05-28", "printed": "64.3%", "where": "benchmark table image, row SWE-Bench Pro, column Opus 4.7" }, { "url": "https://www.minimax.io/blog/minimax-m3", "name": "MiniMax M3", "title": "MiniMax M3: Frontier Coding, 1M Context, Native Multimodality — All in One Model", "lab": "minimax", "date": "2026-06-01", "printed": "64.3", "where": "Bar chart image (10 panels); panel 'SWE Bench Pro'; value label on Opus 4.7 bar" }, { "url": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-5/", "name": "Gemini 3.5 Flash", "title": "Gemini 3.5: frontier intelligence with action", "lab": "google", "date": "2026-05-19", "printed": "64.3%", "where": "Benchmark table image; row 'SWE-Bench Pro (Public), Single attempt'; column claude-opus-4-7" }, { "url": "https://openai.com/index/introducing-gpt-5-5", "name": "GPT-5.5", "title": "Introducing GPT-5.5", "lab": "openai", "date": "2026-04-23", "printed": "64.3%", "where": "table 1 \"SWE-Bench Pro (Public) *\" row, \"Claude Opus 4.7\" column" } ], "source_image": "https://frontierlog.ai/sources/527cd462f89b6de4.png", "source_image_original": "https://www-cdn.anthropic.com/images/4zrzovbb/website/d434d15757c6abac1122af483617741776d5a114-2600x2638.png" }, { "id": "qwen3-6-max-preview:swe-bench-pro:unstated:1", "model": "qwen3-6-max-preview", "model_name": "Qwen3.6-Max-Preview", "model_date": "2026-04-18", "event": "https://frontierlog.ai/event/2026-04-18-qwen3-6-max-preview", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 57.3, "unit": "%", "printed": "57.3", "reported_by": "alibaba", "run_by": null, "label": "Qwen 3.6 Max (preview)", "source": "https://qwen.ai/blog?id=qwen3.6-max-preview", "source_title": "Qwen3.6-Max-Preview: Smarter, Sharper, Still Evolving", "source_date": "2026-04-18", "where": "score chart image under 'Performance' (ten panels), panel 'SWE-bench Pro', first bar (legend 'Qwen 3.6 Max (preview)'), value label", "own": true, "primary": true, "parallel": false, "notes": "", "caveat": "", "human_verified": true, "human_verified_on": "2026-10-19", "check": "human", "corroborated": false, "corroborated_on": null, "corroborated_by": [], "source_image": "https://frontierlog.ai/sources/e01c316dd7a53af0.png", "source_image_original": "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen3.6/Figures/qwen3.6_max_preview_score.png" }, { "id": "kimi-k2-6:swe-bench-pro:unstated:1", "model": "kimi-k2-6", "model_name": "Kimi K2.6", "model_date": "2026-04-20", "event": "https://frontierlog.ai/event/2026-04-20-kimi-k2-6", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 58.6, "unit": "%", "printed": "58.6", "reported_by": "moonshot", "run_by": null, "label": "Kimi K2.6", "source": "https://www.kimi.ai/blog/kimi-k2-6", "source_title": "Kimi K2.6: Advancing Open-Source Coding", "source_date": "2026-04-20", "where": "'Benchmark Table' near the end, row 'SWE-Bench Pro' (Coding), column 'Kimi K2.6' (embedded benchmark JSON; the same value is in the coding chart 'SWE-Bench Pro' near the top). Ignore the 'Model A/Model B' placeholder widget also embedded in the page", "own": true, "primary": true, "parallel": false, "notes": "in-house harness adapted from SWE-agent; thinking mode; average of 10 runs", "caveat": "", "human_verified": true, "human_verified_on": "2026-10-17", "check": "human", "corroborated": true, "corroborated_on": "2026-10-19", "corroborated_by": [ { "url": "https://api-docs.deepseek.com/news/news260424/", "name": "DeepSeek V4 Preview", "title": "DeepSeek V4 Preview Release", "lab": "deepseek", "date": "2026-04-24", "printed": "58.6", "where": "Full benchmark table image; row 'SWE Pro (Resolved)'; column 'K2.6 Thinking'" }, { "url": "https://www.minimax.io/blog/minimax-m3", "name": "MiniMax M3", "title": "MiniMax M3: Frontier Coding, 1M Context, Native Multimodality — All in One Model", "lab": "minimax", "date": "2026-06-01", "printed": "58.6", "where": "Full benchmark table image; row 'SWE-Bench Pro'; column 'Kimi K2.6 Thinking'" }, { "url": "https://thinkingmachines.ai/news/introducing-inkling/", "name": "Inkling", "title": "Inkling: Our Open-Weights Model", "lab": "thinking-machines", "date": "2026-07-15", "printed": "58.6%", "where": "'Benchmarking Inkling' main table, row 'SWEBench Pro Public', column 'Kimi K2.6'" } ], "source_capture": "https://frontierlog.ai/captures/e89624e2a898d0ff.png" }, { "id": "gpt-5-5:swe-bench-pro:unstated:1", "model": "gpt-5-5", "model_name": "GPT-5.5", "model_date": "2026-04-23", "event": "https://frontierlog.ai/event/2026-04-23-gpt-5-5", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 59.4, "unit": "%", "printed": "59.4%", "reported_by": "openai", "run_by": null, "label": "GPT-5.5", "source": "https://openai.com/index/gpt-5-6", "source_title": "GPT-5.6: Frontier intelligence that scales with your ambition", "source_date": "2026-07-09", "where": "coding table (columns GPT-5.6 Sol, Sol Ultra, Terra, Luna, GPT-5.5, ...), row 'SWE-Bench Pro', column 'GPT-5.5'", "own": false, "primary": true, "parallel": false, "notes": "", "caveat": "", "human_verified": true, "human_verified_on": "2026-10-20", "check": "human", "corroborated": false, "corroborated_on": null, "corroborated_by": [], "source_capture": "https://frontierlog.ai/captures/9e2ca9dd2373f9f9.png" }, { "id": "deepseek-v4-pro:swe-bench-pro:unstated:1", "model": "deepseek-v4-pro", "model_name": "DeepSeek-V4-Pro", "model_date": "2026-04-24", "event": "https://frontierlog.ai/event/2026-04-24-deepseek-v4-preview", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 55.4, "unit": "%", "printed": "55.4", "reported_by": "deepseek", "run_by": null, "label": "DS-V4-Pro Max", "source": "https://api-docs.deepseek.com/news/news260424/", "source_title": "DeepSeek V4 Preview Release", "source_date": "2026-04-24", "where": "benchmark table image, row 'SWE Pro (Resolved)' (Agentic), column 'DS-V4-Pro Max'", "own": true, "primary": true, "parallel": false, "notes": "max reasoning effort", "caveat": "", "human_verified": true, "human_verified_on": "2026-10-17", "check": "human", "corroborated": true, "corroborated_on": "2026-10-19", "corroborated_by": [ { "url": "https://www.minimax.io/blog/minimax-m3", "name": "MiniMax M3", "title": "MiniMax M3: Frontier Coding, 1M Context, Native Multimodality — All in One Model", "lab": "minimax", "date": "2026-06-01", "printed": "55.4", "where": "Full benchmark table image; row 'SWE-Bench Pro'; column 'DeepSeek V4 Pro'" }, { "url": "https://thinkingmachines.ai/news/introducing-inkling/", "name": "Inkling", "title": "Inkling: Our Open-Weights Model", "lab": "thinking-machines", "date": "2026-07-15", "printed": "55.4%", "where": "'Benchmarking Inkling' main table, row 'SWEBench Pro Public', column 'DeepSeek V4 Pro'" }, { "url": "https://z.ai/blog/glm-5.2", "name": "GLM-5.2", "title": "GLM-5.2: Built for Long-Horizon Tasks", "lab": "zai", "date": "2026-06-16", "printed": "55.4", "where": "Full Benchmark Table, row 'SWE-bench Pro', column 'DeepSeek-V4-Pro'" } ], "source_image": "https://frontierlog.ai/sources/9226203de8e24b33.png", "source_image_original": "https://api-docs.deepseek.com/img/v4-benchmark-2.png" }, { "id": "laguna-m-1:swe-bench-pro:unstated:1", "model": "laguna-m-1", "model_name": "Laguna M.1", "model_date": "2026-04-28", "event": "https://frontierlog.ai/event/2026-04-28-laguna-xs-2-and-m-1", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 46.9, "unit": "%", "printed": "46.9", "reported_by": "poolside", "run_by": null, "label": "Laguna M.1 (225B-A23B)", "source": "https://poolside.ai/blog/introducing-laguna-xs2-m1", "source_title": "Introducing Laguna XS.2 and Laguna M.1", "source_date": "2026-04-28", "where": "the Laguna M.1 comparison chart ('Our most capable model to date' section), benchmark panel 'SWE-Bench Pro' (Resolved tasks on SWE-Bench Pro), value in the column 'Laguna M.1 (225B-A23B)' (first column; chart rendered by script, read the rendered page)", "own": true, "primary": true, "parallel": false, "notes": "poolside harness on Harbor; pass@1 averaged over 3 runs; some task images patched", "caveat": "", "human_verified": false, "human_verified_on": null, "check": "corroborated", "corroborated": true, "corroborated_on": "2026-11-05", "corroborated_by": [ { "url": "https://poolside.ai/blog/laguna-a-deeper-dive", "name": "Laguna XS.2 and Laguna M.1 card", "title": "Laguna XS.2 and M.1: A Deeper Dive", "lab": "poolside", "date": "2026-04-28", "printed": "46.9", "where": "SWE-Bench Pro bar chart, aria-label 'Laguna M.1 (225B-A23B): 46.9, highlighted', and the HTML table row 'SWE-Bench Pro', Laguna M.1 (225B-A23B) column (Benchmarks as of 28 April 2026)" } ], "source_capture": "https://frontierlog.ai/captures/f69e69cf829ec846.png" }, { "id": "laguna-m-1:swe-bench-pro:unstated:2", "model": "laguna-m-1", "model_name": "Laguna M.1", "model_date": "2026-04-28", "event": "https://frontierlog.ai/event/2026-04-28-laguna-xs-2-and-m-1", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 49.2, "unit": "%", "printed": "49.2", "reported_by": "poolside", "run_by": null, "label": "Laguna M.1 (225B-A23B)", "source": "https://poolside.ai/blog/long-context-update-laguna-xs-2-and-m-1", "source_title": "Long context update: Laguna XS.2 and M.1", "source_date": "2026-05-26", "where": "the single comparison chart under 'Laguna M.1 remains our most capable model', benchmark panel 'SWE-Bench Pro' (Resolved tasks on SWE-Bench Pro), value in the column 'Laguna M.1 (225B-A23B)' (chart rendered by script, read the rendered page)", "own": false, "primary": false, "parallel": false, "notes": "poolside harness on Harbor; pass@1 averaged over 3 runs; some task images patched", "caveat": "", "human_verified": false, "human_verified_on": null, "check": "corroborated", "corroborated": true, "corroborated_on": "2026-11-05", "corroborated_by": [ { "url": "https://huggingface.co/poolside/Laguna-M.1", "name": "Laguna XS.2 and Laguna M.1 card", "title": "Laguna M.1 model card", "lab": "poolside", "date": "2026-04-28", "printed": "49.2%", "where": "Benchmark results table, row '**Laguna M.1**', column 'SWE-bench Pro (Public Dataset)' (README.md)" } ], "source_capture": "https://frontierlog.ai/captures/c192588e44b50c33.png" }, { "id": "laguna-xs-2:swe-bench-pro:unstated:1", "model": "laguna-xs-2", "model_name": "Laguna XS.2", "model_date": "2026-04-28", "event": "https://frontierlog.ai/event/2026-04-28-laguna-xs-2-and-m-1", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 44.5, "unit": "%", "printed": "44.5", "reported_by": "poolside", "run_by": null, "label": "Laguna XS.2 (33B-A3B)", "source": "https://poolside.ai/blog/introducing-laguna-xs2-m1", "source_title": "Introducing Laguna XS.2 and Laguna M.1", "source_date": "2026-04-28", "where": "the Laguna XS.2 comparison chart ('(Extra) Small model, big story' section), benchmark panel 'SWE-Bench Pro' (Resolved tasks on SWE-Bench Pro), value in the column 'Laguna XS.2 (33B-A3B)' (first column; chart rendered by script, read the rendered page)", "own": true, "primary": true, "parallel": false, "notes": "poolside harness on Harbor; pass@1 averaged over 3 runs; some task images patched", "caveat": "", "human_verified": false, "human_verified_on": null, "check": "corroborated", "corroborated": true, "corroborated_on": "2026-11-05", "corroborated_by": [ { "url": "https://poolside.ai/blog/laguna-a-deeper-dive", "name": "Laguna XS.2 and Laguna M.1 card", "title": "Laguna XS.2 and M.1: A Deeper Dive", "lab": "poolside", "date": "2026-04-28", "printed": "44.5", "where": "SWE-Bench Pro bar chart, aria-label 'Laguna XS.2 (33B-A3B): 44.5, highlighted', and the HTML table row 'SWE-Bench Pro', Laguna XS.2 (33B-A3B) column (Benchmarks as of 28 April 2026)" } ], "source_capture": "https://frontierlog.ai/captures/9d496a87332363b8.png" }, { "id": "laguna-xs-2:swe-bench-pro:unstated:2", "model": "laguna-xs-2", "model_name": "Laguna XS.2", "model_date": "2026-04-28", "event": "https://frontierlog.ai/event/2026-04-28-laguna-xs-2-and-m-1", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 46.3, "unit": "%", "printed": "46.3", "reported_by": "poolside", "run_by": null, "label": "Laguna XS.2 (33B-A3B)", "source": "https://poolside.ai/blog/long-context-update-laguna-xs-2-and-m-1", "source_title": "Long context update: Laguna XS.2 and M.1", "source_date": "2026-05-26", "where": "the single comparison chart under 'Laguna M.1 remains our most capable model', benchmark panel 'SWE-Bench Pro' (Resolved tasks on SWE-Bench Pro), value in the column 'Laguna XS.2 (33B-A3B)' (chart rendered by script, read the rendered page)", "own": false, "primary": false, "parallel": false, "notes": "poolside harness on Harbor; pass@1 averaged over 3 runs; some task images patched", "caveat": "", "human_verified": false, "human_verified_on": null, "check": "corroborated", "corroborated": true, "corroborated_on": "2026-11-05", "corroborated_by": [ { "url": "https://huggingface.co/poolside/Laguna-XS.2", "name": "Laguna XS.2 and Laguna M.1 card", "title": "Laguna XS.2 model card", "lab": "poolside", "date": "2026-04-28", "printed": "46.3%", "where": "Benchmark results table, row '**Laguna XS.2**', column 'SWE-bench Pro (Public Dataset)' (README.md)" }, { "url": "https://poolside.ai/blog/introducing-laguna-xs-2-1", "name": "Laguna XS 2.1", "title": "Introducing Laguna XS 2.1", "lab": "poolside", "date": "2026-07-02", "printed": "46.3", "where": "XS 2.1 vs XS.2 bar chart, SWE-Bench Pro, aria-label 'Laguna XS.2 (33B-A3B): 46.3' (comparison bar, Benchmarks as of 2 July 2026)" } ], "source_capture": "https://frontierlog.ai/captures/c192588e44b50c33.png" }, { "id": "claude-opus-4-8:swe-bench-pro:unstated:1", "model": "claude-opus-4-8", "model_name": "Claude Opus 4.8", "model_date": "2026-05-28", "event": "https://frontierlog.ai/event/2026-05-28-claude-opus-4-8", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 69.2, "unit": "%", "printed": "69.2%", "reported_by": "anthropic", "run_by": null, "label": "Opus 4.8", "source": "https://www.anthropic.com/news/claude-opus-4-8", "source_title": "Introducing Claude Opus 4.8", "source_date": "2026-05-28", "where": "benchmark table (image), row 'Agentic coding, SWE-Bench Pro', column 'Opus 4.8'", "own": true, "primary": true, "parallel": false, "notes": "", "caveat": "", "human_verified": true, "human_verified_on": "2026-10-17", "check": "human", "corroborated": true, "corroborated_on": "2026-10-19", "corroborated_by": [ { "url": "https://www.anthropic.com/claude-fable-5-mythos-5-system-card", "name": "Claude Fable 5 and Mythos 5 card", "title": "Claude Fable 5 and Claude Mythos 5 System Card", "lab": "anthropic", "date": "2026-06-09", "printed": "69.2", "where": "Table 8.1.A (p.251), row SWE-bench Pro, column Opus 4.8" }, { "url": "https://www.anthropic.com/claude-opus-4-8-system-card", "name": "Claude Opus 4.8 card", "title": "Claude Opus 4.8 System Card", "lab": "anthropic", "date": "2026-05-28", "printed": "69.2", "where": "Table 8.1.A (p.194), row SWE-bench Pro, column Claude Opus 4.8" }, { "url": "https://www.anthropic.com/claude-opus-5-system-card", "name": "Claude Opus 5 card", "title": "Claude Opus 5 System Card", "lab": "anthropic", "date": "2026-07-24", "printed": "69.2", "where": "Table 8.1.A (p.152), row SWE-bench Pro, column Opus 4.8" }, { "url": "https://www.anthropic.com/news/claude-sonnet-5", "name": "Claude Sonnet 5", "title": "Introducing Claude Sonnet 5", "lab": "anthropic", "date": "2026-06-30", "printed": "69.2%", "where": "benchmark table image, row SWE-bench Pro, column Opus 4.8 (for reference)" }, { "url": "https://openai.com/index/gpt-5-6", "name": "GPT-5.6", "title": "GPT-5.6: Frontier intelligence that scales with your ambition", "lab": "openai", "date": "2026-07-09", "printed": "69.2%", "where": "table 1 \"SWE-Bench Pro\" row, \"Claude Opus 4.8\" column" }, { "url": "https://qwen.ai/blog?id=qwen3.8", "name": "Qwen3.8-Max", "title": "Qwen3.8-Max: A New Bar for Coding and Cowork", "lab": "alibaba", "date": "2026-08-03", "printed": "69.2", "where": "Performance chart image (performance.png), panel SWE-Pro, value label on the Opus4.8 bar" }, { "url": "https://z.ai/blog/glm-5.2", "name": "GLM-5.2", "title": "GLM-5.2: Built for Long-Horizon Tasks", "lab": "zai", "date": "2026-06-16", "printed": "69.2", "where": "Full Benchmark Table, row 'SWE-bench Pro', column 'Claude Opus 4.8'" } ], "source_image": "https://frontierlog.ai/sources/7ad9b784f2579c8a.png", "source_image_original": "https://www-cdn.anthropic.com/images/4zrzovbb/website/a9007019094f217e98cb8261a2765d7646c01708-2600x1392.png" }, { "id": "minimax-m3:swe-bench-pro:unstated:1", "model": "minimax-m3", "model_name": "MiniMax M3", "model_date": "2026-06-01", "event": "https://frontierlog.ai/event/2026-06-01-minimax-m3", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 59.0, "unit": "%", "printed": "59.0", "reported_by": "minimax", "run_by": null, "label": "MiniMax M3", "source": "https://www.minimax.io/blog/minimax-m3", "source_title": "MiniMax M3: Frontier Coding, 1M Context, Native Multimodality — All in One Model", "source_date": "2026-06-01", "where": "benchmark table image, row 'SWE-Bench Pro' (Coding), column 'MiniMax M3' (the post also lists it in prose under the coding paragraph and in the SWE Bench Pro chart image)", "own": true, "primary": true, "parallel": false, "notes": "Claude Code harness", "caveat": "", "human_verified": true, "human_verified_on": "2026-10-17", "check": "human", "corroborated": true, "corroborated_on": "2026-10-19", "corroborated_by": [ { "url": "https://z.ai/blog/glm-5.2", "name": "GLM-5.2", "title": "GLM-5.2: Built for Long-Horizon Tasks", "lab": "zai", "date": "2026-06-16", "printed": "59.0", "where": "Full Benchmark Table, row 'SWE-bench Pro', column 'MiniMax M3'" } ], "source_image": "https://frontierlog.ai/sources/5720f3ad06ea2ce9.jpg", "source_image_original": "https://filecdn.minimax.chat/public/img_v3_02128_b7726cd8-879a-4b7a-a9da-db4395ea597g-1780272508686.jpg" }, { "id": "nemotron-3-ultra-550b-a55b:swe-bench-pro:unstated:1", "model": "nemotron-3-ultra-550b-a55b", "model_name": "Nemotron 3 Ultra 550B-A55B", "model_date": "2026-06-04", "event": "https://frontierlog.ai/event/2026-06-04-nemotron-3-ultra", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 49.66, "unit": "%", "printed": "49.66", "reported_by": "nvidia", "run_by": null, "label": "N3-Ultra 550B-A55B", "source": "https://research.nvidia.com/labs/nemotron/files/NVIDIA-Nemotron-3.5-Super-and-Lightning-Technical-Report.pdf", "source_title": "Nemotron 3.5 Super and Lightning: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Models for Agentic AI", "source_date": "2026-10-15", "where": "Table 10 (PDF page 23), row 'SWE-bench Pro', column 'N3-Ultra 550B-A55B' (second column)", "own": false, "primary": true, "parallel": false, "notes": "OpenCode harness; web search and fetch disabled; average of 3 runs", "caveat": "", "human_verified": false, "human_verified_on": null, "check": "corroborated", "corroborated": true, "corroborated_on": "2026-10-19", "corroborated_by": [ { "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16/blob/1f8953f7f0935358e3605150d22bc15df062061c/README.md", "name": "Nemotron 3.5 Super VL and Bolt card", "title": "nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16 model card (revision of 2026-10-16)", "lab": "nvidia", "date": "2026-10-15", "printed": "49.66", "where": "evaluation table, 'Nemotron 3 Ultra' column, SWE-bench Pro row" }, { "url": "https://research.nvidia.com/labs/nemotron/Nemotron-3.5-Super-and-Lightning/", "name": "Nemotron 3.5 Super VL and Bolt card", "title": "NVIDIA Nemotron 3.5 Super and Lightning", "lab": "nvidia", "date": "2026-10-15", "printed": "49.7", "where": "overview chart, SWE-bench Pro group, 'N3 Ultra 550B-A55B' bar label" } ], "source_capture": "https://frontierlog.ai/captures/ee443a24cf8af62e.png" }, { "id": "claude-fable-5:swe-bench-pro:unstated:1", "model": "claude-fable-5", "model_name": "Claude Fable 5", "model_date": "2026-06-09", "event": "https://frontierlog.ai/event/2026-06-09-claude-fable-5-mythos-5", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 80.0, "unit": "%", "printed": "80", "reported_by": "anthropic", "run_by": null, "label": "Fable 5", "source": "https://www.anthropic.com/claude-fable-5-mythos-5-system-card", "source_title": "Claude Fable 5 and Claude Mythos 5 System Card", "source_date": "2026-06-09", "where": "Section 8.1 Evaluation summary, Table 8.1.A, row 'SWE-bench Pro', column 'Fable 5'", "own": true, "primary": true, "parallel": false, "notes": "adaptive thinking, max effort; average of 5 trials; with production safeguards, including fallback to Opus 4.8", "caveat": "", "human_verified": true, "human_verified_on": "2026-10-17", "check": "human", "corroborated": true, "corroborated_on": "2026-10-19", "corroborated_by": [ { "url": "https://www.anthropic.com/claude-opus-5-system-card", "name": "Claude Opus 5 card", "title": "Claude Opus 5 System Card", "lab": "anthropic", "date": "2026-07-24", "printed": "80", "where": "Table 8.1.A (p.152), row SWE-bench Pro, column Fable 5" }, { "url": "https://openai.com/index/gpt-5-6", "name": "GPT-5.6", "title": "GPT-5.6: Frontier intelligence that scales with your ambition", "lab": "openai", "date": "2026-07-09", "printed": "80%", "where": "table 1 \"SWE-Bench Pro\" row, \"Claude Fable 5\" column" }, { "url": "https://qwen.ai/blog?id=qwen3.8", "name": "Qwen3.8-Max", "title": "Qwen3.8-Max: A New Bar for Coding and Cowork", "lab": "alibaba", "date": "2026-08-03", "printed": "80.0", "where": "Performance chart image (performance.png), panel SWE-Pro, value label on the Fable5 bar" }, { "url": "https://thinkingmachines.ai/news/introducing-inkling/", "name": "Inkling", "title": "Inkling: Our Open-Weights Model", "lab": "thinking-machines", "date": "2026-07-15", "printed": "80.0%", "where": "'Benchmarking Inkling' main table, row 'SWEBench Pro Public', column 'Claude Fable 5 (max)'" } ], "source_capture": "https://frontierlog.ai/captures/b47ed8faf350a8fd.png" }, { "id": "claude-mythos-5:swe-bench-pro:unstated:1", "model": "claude-mythos-5", "model_name": "Claude Mythos 5", "model_date": "2026-06-09", "event": "https://frontierlog.ai/event/2026-06-09-claude-fable-5-mythos-5", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 80.3, "unit": "%", "printed": "80.3", "reported_by": "anthropic", "run_by": null, "label": "Mythos 5", "source": "https://www.anthropic.com/claude-fable-5-mythos-5-system-card", "source_title": "Claude Fable 5 and Claude Mythos 5 System Card", "source_date": "2026-06-09", "where": "Section 8.1 Evaluation summary, Table 8.1.A, row 'SWE-bench Pro', column 'Mythos 5'", "own": true, "primary": true, "parallel": false, "notes": "adaptive thinking, max effort; average of 5 trials", "caveat": "", "human_verified": true, "human_verified_on": "2026-10-17", "check": "human", "corroborated": true, "corroborated_on": "2026-10-19", "corroborated_by": [ { "url": "https://openai.com/index/gpt-5-6", "name": "GPT-5.6", "title": "GPT-5.6: Frontier intelligence that scales with your ambition", "lab": "openai", "date": "2026-07-09", "printed": "80.3%", "where": "table 1 \"SWE-Bench Pro\" row, \"Claude Mythos 5\" column" } ], "source_capture": "https://frontierlog.ai/captures/b47ed8faf350a8fd.png" }, { "id": "glm-5-2:swe-bench-pro:unstated:1", "model": "glm-5-2", "model_name": "GLM-5.2", "model_date": "2026-06-16", "event": "https://frontierlog.ai/event/2026-06-16-glm-5-2", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 62.1, "unit": "%", "printed": "62.1", "reported_by": "zai", "run_by": null, "label": "GLM-5.2", "source": "https://z.ai/blog/glm-5.2", "source_title": "GLM-5.2: Built for Long-Horizon Tasks", "source_date": "2026-06-16", "where": "'Full Benchmark Table', row 'SWE-bench Pro', column 'GLM-5.2'", "own": true, "primary": true, "parallel": false, "notes": "OpenHands with a tailored prompt; 400K context", "caveat": "", "human_verified": true, "human_verified_on": "2026-10-19", "check": "human", "corroborated": true, "corroborated_on": "2026-10-19", "corroborated_by": [ { "url": "https://thinkingmachines.ai/news/introducing-inkling/", "name": "Inkling", "title": "Inkling: Our Open-Weights Model", "lab": "thinking-machines", "date": "2026-07-15", "printed": "62.1%", "where": "'Benchmarking Inkling' main table, row 'SWEBench Pro Public', column 'GLM 5.2'" } ], "source_capture": "https://frontierlog.ai/captures/112d3aba81edfc46.png" }, { "id": "claude-sonnet-5:swe-bench-pro:unstated:1", "model": "claude-sonnet-5", "model_name": "Claude Sonnet 5", "model_date": "2026-06-30", "event": "https://frontierlog.ai/event/2026-06-30-claude-sonnet-5", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 63.2, "unit": "%", "printed": "63.2%", "reported_by": "anthropic", "run_by": null, "label": "Sonnet 5", "source": "https://www.anthropic.com/news/claude-sonnet-5", "source_title": "Introducing Claude Sonnet 5", "source_date": "2026-06-30", "where": "benchmark table (image), row 'Agentic coding, SWE-bench Pro', column 'Sonnet 5'", "own": true, "primary": true, "parallel": false, "notes": "", "caveat": "", "human_verified": true, "human_verified_on": "2026-10-17", "check": "human", "corroborated": true, "corroborated_on": "2026-10-19", "corroborated_by": [ { "url": "https://www.anthropic.com/claude-sonnet-5-5-system-card", "name": "Claude Sonnet 5.5 card", "title": "Claude Sonnet 5.5 System Card", "lab": "anthropic", "date": "2026-09-28", "printed": "63.2", "where": "Table 8.1.A (p.109), row SWE-Bench Pro, column Claude Sonnet 5" }, { "url": "https://www.anthropic.com/claude-sonnet-5-system-card", "name": "Claude Sonnet 5 card", "title": "Claude Sonnet 5 System Card", "lab": "anthropic", "date": "2026-06-30", "printed": "63.2", "where": "Table 8.1.A (p.115), row SWE-bench Pro, column Claude Sonnet 5" }, { "url": "https://deepmind.google/models/model-cards/gemini-3-6-flash/", "name": "Gemini 3.6 Flash and 3.5 Flash-Lite card", "title": "Gemini 3.6 Flash - Model Card", "lab": "google", "date": "2026-07-21", "printed": "63.2%", "where": "HTML results table 'Results as of July, 2026'; row 'SWE-Bench Pro (Public)'; column claude-sonnet-5" } ], "source_image": "https://frontierlog.ai/sources/849d6a908f0d46ac.png", "source_image_original": "https://www-cdn.anthropic.com/images/4zrzovbb/website/9941d610909f28a504e16dd5af823df172ec6035-2600x1234.png" }, { "id": "laguna-xs-2-1:swe-bench-pro:unstated:1", "model": "laguna-xs-2-1", "model_name": "Laguna XS 2.1", "model_date": "2026-07-02", "event": "https://frontierlog.ai/event/2026-07-02-laguna-xs-2-1", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 47.6, "unit": "%", "printed": "47.6", "reported_by": "poolside", "run_by": null, "label": "Laguna XS 2.1 (33B-A3B)", "source": "https://poolside.ai/blog/introducing-laguna-xs-2-1", "source_title": "Introducing Laguna XS 2.1", "source_date": "2026-07-02", "where": "evaluation chart 'SWE-Bench Pro' (Resolved tasks on SWE-Bench Pro), bar 'Laguna XS 2.1 (33B-A3B)', value label (animates up from 0.0 when scrolled into view; also in the bar's aria-label)", "own": true, "primary": true, "parallel": false, "notes": "poolside harness on Harbor; pass@1 averaged over 2 attempts; some task images patched", "caveat": "", "human_verified": true, "human_verified_on": "2026-10-17", "check": "human", "corroborated": false, "corroborated_on": null, "corroborated_by": [], "source_capture": "https://frontierlog.ai/captures/c9c98769b50ed413.png" }, { "id": "hy3:swe-bench-pro:unstated:1", "model": "hy3", "model_name": "Hy3", "model_date": "2026-07-06", "event": "https://frontierlog.ai/event/2026-07-06-hy3", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 57.9, "unit": "%", "printed": "57.9", "reported_by": "tencent", "run_by": null, "label": "Hy3", "source": "https://hy.tencent.ai/research/hy3", "source_title": "Introducing Hy3", "source_date": "2026-07-06", "where": "'Appendix: Benchmark Results' table image, row 'SWE-bench Pro', column 'Hy3'", "own": true, "primary": true, "parallel": false, "notes": "SWE-agent scaffold; highest reasoning setting", "caveat": "", "human_verified": false, "human_verified_on": null, "check": "corroborated", "corroborated": true, "corroborated_on": "2026-11-05", "corroborated_by": [ { "url": "https://huggingface.co/tencent/Hy3", "name": "Hy3 card", "title": "tencent/Hy3 model card", "lab": "tencent", "date": "2026-07-06", "printed": "57.9", "where": "Model card, Appendix image, row SWE-bench Pro, column Hy3. Same table as the Hy3 page's appendix, but a different image file (the page's is d1af768b…). The card's top bar chart has the same bytes as the page's chart, so it is not cited. Mirrored with identical bytes on huggingface.co/tencent/Hy3-FP8 and github.com/Tencent-Hunyuan/Hy3 (assets/benchmark-appendix.png). Notes: highest reasoning tier; SWE-agent scaffold; Terminus-2, 4h timeout, 500 episodes; MCP-Atlas 500-task public set, 100-call budget, Gemini 2.5 Pro judge; BrowseComp self-summary context management." }, { "url": "https://hy.tencent.ai/research/hy4-preview", "name": "Hy4 preview", "title": "Introducing Hy4 preview", "lab": "tencent", "date": "2026-08-28", "printed": "57.9", "where": "Appendix: Benchmark Results image, row SWE-bench Pro, column Hy3 (swe-agent scaffold, highest reasoning setting). Hy4 preview appendix (Aug 28, 2026), Hy3 column. Notes say 'Certain Hy3 benchmark scores may differ from previously reported results following updates to the evaluation harness, judge model, and anti-hacking mechanisms.' Same image bytes as huggingface.co/tencent/Hy4-preview assets/benchmark-appendix.jpg and github.com/Tencent-Hunyuan/Hy4-preview." } ], "source_image": "https://frontierlog.ai/sources/d1af768b6734a711.png", "source_image_original": "https://hy-model-ap-prod-1258344703.cos.accelerate.myqcloud.com/llm-blog/default/9a5ea02f/2026070622113393_004bbe5190853d7c35825b9257672dde.png?q-sign-algorithm=sha1&q-ak=AKIDRl074nOsGdJ9zjMsCRWP3ShmgS3VtX4S&q-sign-time=1783347098;2098707098&q-key-time=1783347098;2098707098&q-header-list=&q-url-param-list=&q-signature=507857321cef2b9fef5e4ef8a8f9ab8937d2480b" }, { "id": "hy3:swe-bench-pro:unstated:2", "model": "hy3", "model_name": "Hy3", "model_date": "2026-07-06", "event": "https://frontierlog.ai/event/2026-07-06-hy3", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 46.0, "unit": "%", "printed": "46.0", "reported_by": "tencent", "run_by": null, "label": "Hy3 preview", "source": "https://hy.tencent.ai/research/hy3", "source_title": "Introducing Hy3", "source_date": "2026-07-06", "where": "'Appendix: Benchmark Results' table image, row 'SWE-bench Pro', column 'Hy3 preview'", "own": true, "primary": false, "parallel": false, "notes": "Hy3 preview, re-evaluated at Hy3 launch; SWE-agent scaffold; highest reasoning setting", "caveat": "", "human_verified": false, "human_verified_on": null, "check": "corroborated", "corroborated": true, "corroborated_on": "2026-11-05", "corroborated_by": [ { "url": "https://huggingface.co/tencent/Hy3", "name": "Hy3 card", "title": "tencent/Hy3 model card", "lab": "tencent", "date": "2026-07-06", "printed": "46.0", "where": "Model card, Appendix image, row SWE-bench Pro, column Hy3 preview. Same table as the Hy3 page's appendix, but a different image file (the page's is d1af768b…). The card's top bar chart has the same bytes as the page's chart, so it is not cited. Mirrored with identical bytes on huggingface.co/tencent/Hy3-FP8 and github.com/Tencent-Hunyuan/Hy3 (assets/benchmark-appendix.png). Notes: highest reasoning tier; SWE-agent scaffold; Terminus-2, 4h timeout, 500 episodes; MCP-Atlas 500-task public set, 100-call budget, Gemini 2.5 Pro judge; BrowseComp self-summary context management." }, { "url": "https://hy.tencent.ai/research/hy3-preview", "name": "Tencent releases Hy3 preview", "title": "Hy3 preview: The First Step in Rebuilding the Hy model", "lab": "tencent", "date": "2026-04-23", "printed": "46.0", "where": "Appendix table image, row SWE-bench Pro, column Hy3 preview. Hy3 preview launch page (Apr 23, 2026), appendix table, 'Hy3 preview' column. Notes: competitors at high-think; 'For the SWE-Bench series ... SWE-agent scaffold, with the best results obtained in non-thinking mode.'" } ], "source_image": "https://frontierlog.ai/sources/d1af768b6734a711.png", "source_image_original": "https://hy-model-ap-prod-1258344703.cos.accelerate.myqcloud.com/llm-blog/default/9a5ea02f/2026070622113393_004bbe5190853d7c35825b9257672dde.png?q-sign-algorithm=sha1&q-ak=AKIDRl074nOsGdJ9zjMsCRWP3ShmgS3VtX4S&q-sign-time=1783347098;2098707098&q-key-time=1783347098;2098707098&q-header-list=&q-url-param-list=&q-signature=507857321cef2b9fef5e4ef8a8f9ab8937d2480b" }, { "id": "gpt-5-6-luna:swe-bench-pro:unstated:1", "model": "gpt-5-6-luna", "model_name": "GPT-5.6 Luna", "model_date": "2026-07-09", "event": "https://frontierlog.ai/event/2026-07-09-gpt-5-6", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 62.7, "unit": "%", "printed": "62.7%", "reported_by": "openai", "run_by": null, "label": "GPT‑5.6 Luna", "source": "https://openai.com/index/gpt-5-6", "source_title": "GPT-5.6: Frontier intelligence that scales with your ambition", "source_date": "2026-07-09", "where": "coding eval table (columns GPT‑5.6 Sol, Sol Ultra, Terra, Luna, GPT‑5.5, …), row 'SWE-Bench Pro', column 'GPT‑5.6 Luna'", "own": true, "primary": true, "parallel": false, "notes": "", "caveat": "", "human_verified": true, "human_verified_on": "2026-10-26", "check": "human", "corroborated": false, "corroborated_on": null, "corroborated_by": [], "source_capture": "https://frontierlog.ai/captures/9e2ca9dd2373f9f9.png" }, { "id": "gpt-5-6-sol:swe-bench-pro:unstated:1", "model": "gpt-5-6-sol", "model_name": "GPT-5.6 Sol", "model_date": "2026-07-09", "event": "https://frontierlog.ai/event/2026-07-09-gpt-5-6", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 64.6, "unit": "%", "printed": "64.6%", "reported_by": "openai", "run_by": null, "label": "GPT‑5.6 Sol", "source": "https://openai.com/index/gpt-5-6", "source_title": "GPT-5.6: Frontier intelligence that scales with your ambition", "source_date": "2026-07-09", "where": "coding eval table (columns GPT‑5.6 Sol, Sol Ultra, Terra, Luna, GPT‑5.5, …), row 'SWE-Bench Pro', column 'GPT‑5.6 Sol'", "own": true, "primary": true, "parallel": false, "notes": "", "caveat": "", "human_verified": true, "human_verified_on": "2026-10-17", "check": "human", "corroborated": true, "corroborated_on": "2026-10-19", "corroborated_by": [ { "url": "https://www.anthropic.com/claude-fable-5-1-mythos-5-1-system-card", "name": "Claude Fable 5.1 and Mythos 5.1 card", "title": "Claude Fable 5.1 and Claude Mythos 5.1 System Card", "lab": "anthropic", "date": "2026-09-01", "printed": "64.6", "where": "Table 8.1.A (p.167), row SWE-bench Pro, column GPT-5.6 Sol" }, { "url": "https://www.anthropic.com/claude-opus-5-system-card", "name": "Claude Opus 5 card", "title": "Claude Opus 5 System Card", "lab": "anthropic", "date": "2026-07-24", "printed": "64.6", "where": "Table 8.1.A (p.152), row SWE-bench Pro, column GPT 5.6 Sol" }, { "url": "https://qwen.ai/blog?id=qwen3.8", "name": "Qwen3.8-Max", "title": "Qwen3.8-Max: A New Bar for Coding and Cowork", "lab": "alibaba", "date": "2026-08-03", "printed": "64.6", "where": "Performance chart image (performance.png), panel SWE-Pro, value label on the GPT5.6 Sol (max) bar" }, { "url": "https://thinkingmachines.ai/news/introducing-inkling/", "name": "Inkling", "title": "Inkling: Our Open-Weights Model", "lab": "thinking-machines", "date": "2026-07-15", "printed": "64.6%", "where": "'Benchmarking Inkling' main table, row 'SWEBench Pro Public', column 'GPT 5.6 Sol (max/xhigh)'" } ], "source_capture": "https://frontierlog.ai/captures/9e2ca9dd2373f9f9.png" }, { "id": "gpt-5-6-terra:swe-bench-pro:unstated:1", "model": "gpt-5-6-terra", "model_name": "GPT-5.6 Terra", "model_date": "2026-07-09", "event": "https://frontierlog.ai/event/2026-07-09-gpt-5-6", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 63.4, "unit": "%", "printed": "63.4%", "reported_by": "openai", "run_by": null, "label": "GPT‑5.6 Terra", "source": "https://openai.com/index/gpt-5-6", "source_title": "GPT-5.6: Frontier intelligence that scales with your ambition", "source_date": "2026-07-09", "where": "coding eval table (columns GPT‑5.6 Sol, Sol Ultra, Terra, Luna, GPT‑5.5, …), row 'SWE-Bench Pro', column 'GPT‑5.6 Terra'", "own": true, "primary": true, "parallel": false, "notes": "", "caveat": "", "human_verified": true, "human_verified_on": "2026-10-26", "check": "human", "corroborated": false, "corroborated_on": null, "corroborated_by": [], "source_capture": "https://frontierlog.ai/captures/9e2ca9dd2373f9f9.png" }, { "id": "claude-opus-5:swe-bench-pro:unstated:1", "model": "claude-opus-5", "model_name": "Claude Opus 5", "model_date": "2026-07-24", "event": "https://frontierlog.ai/event/2026-07-24-claude-opus-5", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 79.2, "unit": "%", "printed": "79.2", "reported_by": "anthropic", "run_by": null, "label": "Opus 5", "source": "https://www.anthropic.com/claude-opus-5-system-card", "source_title": "Claude Opus 5 System Card", "source_date": "2026-07-24", "where": "Section 8.1 Evaluation summary, Table 8.1.A, row 'SWE-bench Pro', column 'Opus 5'", "own": true, "primary": true, "parallel": false, "notes": "adaptive thinking, max effort; average of 5 trials", "caveat": "", "human_verified": true, "human_verified_on": "2026-10-17", "check": "human", "corroborated": true, "corroborated_on": "2026-10-19", "corroborated_by": [ { "url": "https://www.anthropic.com/claude-fable-5-1-mythos-5-1-system-card", "name": "Claude Fable 5.1 and Mythos 5.1 card", "title": "Claude Fable 5.1 and Claude Mythos 5.1 System Card", "lab": "anthropic", "date": "2026-09-01", "printed": "79.2", "where": "Table 8.1.A (p.167), row SWE-bench Pro, column Claude Opus 5" }, { "url": "https://www.anthropic.com/claude-opus-5-5-system-card", "name": "Claude Opus 5.5 card", "title": "Claude Opus 5.5 System Card", "lab": "anthropic", "date": "2026-09-22", "printed": "79.2", "where": "Table 8.1.A (p.174), row SWE-bench Pro, column Claude Opus 5" } ], "source_capture": "https://frontierlog.ai/captures/057c8f7b933f9831.png" }, { "id": "hy4-preview:swe-bench-pro:unstated:1", "model": "hy4-preview", "model_name": "Hy4 preview", "model_date": "2026-08-28", "event": "https://frontierlog.ai/event/2026-08-28-hy4-preview", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 65.7, "unit": "%", "printed": "65.7", "reported_by": "tencent", "run_by": null, "label": "Hy4 preview", "source": "https://hy.tencent.ai/research/hy4-preview", "source_title": "Introducing Hy4 preview", "source_date": "2026-08-28", "where": "'Appendix: Benchmark Results' table image, row 'SWE-bench Pro', column 'Hy4 preview'", "own": true, "primary": true, "parallel": false, "notes": "SWE-agent scaffold; highest reasoning setting", "caveat": "", "human_verified": true, "human_verified_on": "2026-11-09", "check": "human", "corroborated": false, "corroborated_on": null, "corroborated_by": [], "source_image": "https://frontierlog.ai/sources/02c751bf51c52c52.jpg", "source_image_original": "https://hy-model-ap-prod-1258344703.cos.accelerate.myqcloud.com/llm-blog/default/9a5ea02f/2026082812342438_ecb98509e92e4be02a40ca358414133a.jpg?q-sign-algorithm=sha1&q-ak=AKIDRl074nOsGdJ9zjMsCRWP3ShmgS3VtX4S&q-sign-time=1787891670;2103251670&q-key-time=1787891670;2103251670&q-header-list=&q-url-param-list=&q-signature=b6d0c3c4d6d7e0e5219d177ec730735185ccdd1f" }, { "id": "claude-fable-5-1:swe-bench-pro:unstated:1", "model": "claude-fable-5-1", "model_name": "Claude Fable 5.1", "model_date": "2026-09-01", "event": "https://frontierlog.ai/event/2026-09-01-claude-fable-5-1-mythos-5-1", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 81.2, "unit": "%", "printed": "81.2%", "reported_by": "anthropic", "run_by": null, "label": "Fable 5.1", "source": "https://www.anthropic.com/claude-fable-5-1-mythos-5-1-system-card", "source_title": "Claude Fable 5.1 and Claude Mythos 5.1 System Card", "source_date": "2026-09-01", "where": "Section 8.2 'SWE-bench Pro, Multilingual, and Multimodal', first bullet", "own": true, "primary": true, "parallel": false, "notes": "average of 5 trials", "caveat": "", "human_verified": true, "human_verified_on": "2026-10-17", "check": "human", "corroborated": true, "corroborated_on": "2026-10-19", "corroborated_by": [ { "url": "https://www.anthropic.com/claude-opus-5-5-system-card", "name": "Claude Opus 5.5 card", "title": "Claude Opus 5.5 System Card", "lab": "anthropic", "date": "2026-09-22", "printed": "81.2", "where": "Table 8.1.A (p.174), row SWE-bench Pro, column Claude Fable 5.1" } ] }, { "id": "claude-opus-5-5:swe-bench-pro:unstated:1", "model": "claude-opus-5-5", "model_name": "Claude Opus 5.5", "model_date": "2026-09-22", "event": "https://frontierlog.ai/event/2026-09-22-claude-opus-5-5", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 89.9, "unit": "%", "printed": "89.9", "reported_by": "anthropic", "run_by": null, "label": "Claude Opus 5.5", "source": "https://www.anthropic.com/claude-opus-5-5-system-card", "source_title": "Claude Opus 5.5 System Card", "source_date": "2026-09-22", "where": "Section 8.1 Evaluation summary, Table 8.1.A, row 'SWE-bench Pro', column 'Claude Opus 5.5'", "own": true, "primary": true, "parallel": false, "notes": "adaptive thinking, max effort; average of 5 trials", "caveat": "", "human_verified": true, "human_verified_on": "2026-10-17", "check": "human", "corroborated": true, "corroborated_on": "2026-10-19", "corroborated_by": [ { "url": "https://www.anthropic.com/claude-sonnet-5-5-system-card", "name": "Claude Sonnet 5.5 card", "title": "Claude Sonnet 5.5 System Card", "lab": "anthropic", "date": "2026-09-28", "printed": "89.9", "where": "Table 8.1.A (p.109), row SWE-Bench Pro, column Claude Opus 5.5" } ], "source_capture": "https://frontierlog.ai/captures/8a49bd8c526c6c66.png" }, { "id": "claude-sonnet-5-5:swe-bench-pro:unstated:1", "model": "claude-sonnet-5-5", "model_name": "Claude Sonnet 5.5", "model_date": "2026-09-28", "event": "https://frontierlog.ai/event/2026-09-28-claude-sonnet-5-5", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 81.3, "unit": "%", "printed": "81.3", "reported_by": "anthropic", "run_by": null, "label": "Claude Sonnet 5.5", "source": "https://www.anthropic.com/claude-sonnet-5-5-system-card", "source_title": "Claude Sonnet 5.5 System Card", "source_date": "2026-09-28", "where": "Section 8.1 Evaluation summary, Table 8.1.A, row 'SWE-Bench Pro', column 'Claude Sonnet 5.5'", "own": true, "primary": true, "parallel": false, "notes": "adaptive thinking, max effort; average of 5 trials", "caveat": "", "human_verified": true, "human_verified_on": "2026-10-17", "check": "human", "corroborated": true, "corroborated_on": "2026-10-19", "corroborated_by": [ { "url": "https://www.anthropic.com/claude-haiku-5-5-system-card", "name": "Claude Haiku 5.5 card", "title": "Claude Haiku 5.5 System Card", "lab": "anthropic", "date": "2026-10-07", "printed": "81.3", "where": "Table 8.1.A (p.111), row SWE-Bench Pro, column Sonnet 5.5" } ], "source_capture": "https://frontierlog.ai/captures/53b68efd4b35b2e3.png" }, { "id": "claude-haiku-5-5:swe-bench-pro:unstated:1", "model": "claude-haiku-5-5", "model_name": "Claude Haiku 5.5", "model_date": "2026-10-07", "event": "https://frontierlog.ai/event/2026-10-07-claude-haiku-5-5", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 64.8, "unit": "%", "printed": "64.8", "reported_by": "anthropic", "run_by": null, "label": "Haiku 5.5", "source": "https://www.anthropic.com/claude-haiku-5-5-system-card", "source_title": "Claude Haiku 5.5 System Card", "source_date": "2026-10-07", "where": "Section 8.1 Evaluation summary, Table 8.1.A, row 'SWE-Bench Pro', column 'Haiku 5.5'", "own": true, "primary": true, "parallel": false, "notes": "adaptive thinking, max effort; average of 5 trials", "caveat": "", "human_verified": true, "human_verified_on": "2026-10-17", "check": "human", "corroborated": false, "corroborated_on": null, "corroborated_by": [], "source_capture": "https://frontierlog.ai/captures/1189b487a0e11054.png" }, { "id": "nemotron-3-5-super-vl-120b-a12b:swe-bench-pro:unstated:1", "model": "nemotron-3-5-super-vl-120b-a12b", "model_name": "Nemotron 3.5 Super VL 120B-A12B", "model_date": "2026-10-15", "event": "https://frontierlog.ai/event/2026-10-15-nemotron-3-5-super-vl-bolt", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 51.78, "unit": "%", "printed": "51.78", "reported_by": "nvidia", "run_by": null, "label": "Nemotron 3.5 Super VL BF16", "source": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-NVFP4/blob/8a115b9d966adce8c059a7bcf47f49a1e57957d8/README.md", "source_title": "nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-NVFP4 model card (launch-day revision)", "source_date": "2026-10-15", "where": "'Benchmarks' table, row 'SWE-bench Pro', column 'Nemotron 3.5 Super VL BF16'", "own": true, "primary": true, "parallel": false, "notes": "", "caveat": "", "human_verified": false, "human_verified_on": null, "check": "corroborated", "corroborated": true, "corroborated_on": "2026-10-19", "corroborated_by": [ { "url": "https://research.nvidia.com/labs/nemotron/files/NVIDIA-Nemotron-3.5-Super-and-Lightning-Technical-Report.pdf", "name": "Nemotron 3.5 Super VL and Bolt card", "title": "Nemotron 3.5 Super and Lightning: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Models for Agentic AI", "lab": "nvidia", "date": "2026-10-15", "printed": "51.78", "where": "Table 10 (Evaluation suite for Nemotron 3.5 Super), SWE-bench Pro row, N3.5-Super column" }, { "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16/blob/1f8953f7f0935358e3605150d22bc15df062061c/README.md", "name": "Nemotron 3.5 Super VL and Bolt card", "title": "nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16 model card (revision of 2026-10-16)", "lab": "nvidia", "date": "2026-10-15", "printed": "51.78", "where": "evaluation table, 'Nemotron 3.5 SuperGA BF16' column, SWE-bench Pro row" }, { "url": "https://research.nvidia.com/labs/nemotron/Nemotron-3.5-Super-and-Lightning/", "name": "Nemotron 3.5 Super VL and Bolt card", "title": "NVIDIA Nemotron 3.5 Super and Lightning", "lab": "nvidia", "date": "2026-10-15", "printed": "51.78%", "where": "Highlights: 'Nemotron 3.5 Super achieves 51.78% on SWE-bench Pro'" } ], "source_capture": "https://frontierlog.ai/captures/86e9dc24de0f0cf9.png" }, { "id": "ripple-65b-preview:swe-bench-pro:unstated:1", "model": "ripple-65b-preview", "model_name": "Ripple-65B Preview", "model_date": "2026-11-04", "event": "https://frontierlog.ai/event/2026-11-04-ripple-65b-preview", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 61.42, "unit": "%", "printed": "61.42", "reported_by": "ricursive", "run_by": null, "label": "Ripple-65B Preview", "source": "https://huggingface.co/ricursiveAI/Ripple-65B-Preview", "source_title": "ricursiveAI/Ripple-65B-Preview model card", "source_date": "2026-11-04", "where": "Card, 'Instruct Model Benchmarks' table, row 'SWE-bench Pro v1.0', column 'Ripple-65B Preview'", "own": true, "primary": true, "parallel": false, "notes": "v1.0; OpenCode harness; maximum effort", "caveat": "", "human_verified": true, "human_verified_on": "2026-11-07", "check": "human", "corroborated": false, "corroborated_on": null, "corroborated_by": [], "source_capture": "https://frontierlog.ai/captures/b689899dbfbbdf93.png" }, { "id": "claude-fable-5-2:swe-bench-pro:unstated:1", "model": "claude-fable-5-2", "model_name": "Claude Fable 5.2", "model_date": "2026-11-09", "event": "https://frontierlog.ai/event/2026-11-09-claude-fable-5-2-and-mythos-5-2", "benchmark": "swe-bench-pro", "benchmark_name": "SWE-bench Pro", "setting": "unstated", "setting_name": "Set not stated", "score": 93.0, "unit": "%", "printed": "93.0", "reported_by": "anthropic", "run_by": null, "label": "Fable 5.2", "source": "https://www.anthropic.com/claude-fable-5-2-mythos-5-2-system-card", "source_title": "Claude Fable 5.2 and Claude Mythos 5.2 System Card", "source_date": "2026-11-09", "where": "Section 8.1 Evaluation summary, Table 8.1.A, row 'SWE-bench, Pro', column 'Fable 5.2'", "own": true, "primary": true, "parallel": false, "notes": "adaptive thinking, max effort; average of 5 trials", "caveat": "", "human_verified": false, "human_verified_on": null, "check": "double_entry", "corroborated": false, "corroborated_on": null, "corroborated_by": [] } ] }