LiveCodeBench v6
Created by UC Berkeley, MIT and Cornell, 21 Apr 2025 · Higher is better · Website
- Other labs
- Reported in a later post or card
Results
| Model | Score | Source | Source check |
|---|---|---|---|
| Nemotron 3.5 Super VL 120B-A12B15 Oct 2026 · NVIDIA | 88.22% | nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16 model card (revision of 2026-10-15)Our capture | |
| Qwen3.8-Omni-Flash18 Sep 2026 · Alibaba | 92.6% | Qwen3.8-Omni-Flash: Omni Senses. Agentic Delivery.Our capture | CheckedA person checked this number against its source on 28 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| Qwen3.8-Flash-Next26 Aug 2026 · Alibaba | 91.9% | Qwen3.8-Flash-Next: A New Architecture, Towards Ultimate Cost-EfficiencyOur capture | CheckedA person checked this number against its source on 28 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.CorroboratedThe same number is printed in 1 other lab document: 1 |
| DiffusionGemma 26B A4B10 Jun 2026 · Google | 69.1% | DiffusionGemma: 4x faster text generationSource image | CorroboratedThe same number is printed in 1 other lab document: 1 |
| Nemotron 3 Ultra 550B-A55B4 Jun 2026 · NVIDIA | 89.0% | Nemotron 3 Ultra: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Model for Agentic ReasoningOur capture | CorroboratedThe same number is printed in 1 other lab document: 1 |
| Nemotron 3 Ultra 550B-A55B4 Jun 2026 · NVIDIA | 89.21% | Nemotron 3.5 Super and Lightning: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Models for Agentic AIproblems from Aug 2024 to May 2025; average of 3 samples; reported later; another report, not chartedOur capture | CorroboratedThe same number is printed in 1 other lab document: 1 |
| Gemma 4 12B3 Jun 2026 · Google | 72% | Introducing Gemma 4 12B: a unified, encoder-free multimodal modelSource image | |
| Qwen3.7-Plus1 Jun 2026 · Alibaba | 89.6% | Qwen3.7-Plus: Multimodal Agent IntelligenceOur capture | CheckedA person checked this number against its source on 28 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.CorroboratedThe same number is printed in 1 other lab document: 1 |
| Kimi K2.620 Apr 2026 · Moonshot AI | 89.6% | Kimi K2.6: Advancing Open-Source CodingOur capture | CorroboratedThe same number is printed in 1 other lab document: 1 |
| Qwen3.6-35B-A3B15 Apr 2026 · Alibaba | 80.4% | Qwen3.6-35B-A3B: Agentic Coding Power, Now Open to AllOur capture | CheckedA person checked this number against its source on 28 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.CorroboratedThe same number is printed in 1 other lab document: 1 |
| Gemma 4 31B2 Apr 2026 · Google | 80.0% | Gemma 4: Byte for byte, the most capable open modelsthinkingSource image | |
| Qwen3.6-Plus2 Apr 2026 · Alibaba | 87.1% | Qwen3.6-Plus: Towards Real World AgentsOur capture | CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| Qwen3.5-Omni30 Mar 2026 · Alibaba | 65.6% | Qwen3.5-Omni: Scaling Up, Toward Native Omni-Modal AGIQwen3.5-Omni-Plus; no thinkingOur capture | CheckedA person checked this number against its source on 26 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| Qwen3.5-Omni-Flash30 Mar 2026 · Alibaba | 56.6% | Qwen3.5-Omni: Scaling Up, Toward Native Omni-Modal AGIQwen3.5-Omni-Flash; no thinkingOur capture | CheckedA person checked this number against its source on 28 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| Nemotron 3 Super 120B-A12B11 Mar 2026 · NVIDIA | 78.69% | nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8 model card (launch-day revision)problems from Aug 2024 to May 2025Our capture | |
| Qwen3.5-397B-A17B16 Feb 2026 · Alibaba | 83.6% | Qwen3.5: Towards Native Multimodal AgentsOur capture | CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| Kimi K2.527 Jan 2026 · Moonshot AI | 85.0% | Kimi K2.5: Visual Agentic Intelligencethinking mode; average of 5 runsOur capture | |
| Qwen3-Max-Thinking26 Jan 2026 · Alibaba | 85.9% | Pushing Qwen3-Max-Thinking Beyond its Limitsproblems from Feb to May 2025Our capture | CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| Qwen3-Max-Thinking26 Jan 2026 · Alibaba | 91.4% | Pushing Qwen3-Max-Thinking Beyond its Limitsheavy mode: test-time scaling over parallel trajectories; parallel test-time computeSource image | CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| Qwen3-Max-Thinking26 Jan 2026 · Alibaba | 88.0% | Pushing Qwen3-Max-Thinking Beyond its Limitsstandard parallel sampling and aggregation, the baseline for heavy mode; parallel test-time compute | CheckedA person checked this number against its source on 28 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| GLM-4.722 Dec 2025 · Z.ai | 84.9% | GLM-4.7: Advancing the Coding CapabilityOur capture | CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| Nemotron 3 Nano 30B-A3B15 Dec 2025 · NVIDIA | 68.3% | nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-FP8 model card (launch-day revision)Our capture | |
| Nemotron 3 Nano 30B-A3B15 Dec 2025 · NVIDIA | 68.2% | NVIDIA Nemotron 3 Family of Modelsanother report, not chartedSource image | CheckedA person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| Kimi K2 Thinking6 Nov 2025 · Moonshot AI | 83.1% | Introducing Kimi K2 Thinking128K thinking budget; INT4Our capture | |
| GLM-4.630 Sep 2025 · Z.ai | 82.8% | GLM-4.6: Advanced Agentic, Reasoning and Coding Capabilitiesno tools; 128K contextSource image | CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.CorroboratedThe same number is printed in 1 other lab document: 1 |
| GLM-4.630 Sep 2025 · Z.ai | 84.5% | GLM-4.6: Advanced Agentic, Reasoning and Coding Capabilitieswith tools; 128K context; another report, not chartedSource image | CorroboratedThe same number is printed in 1 other lab document: 1 |
| Qwen3-Max24 Sep 2025 · Alibaba | 69.0% | Qwen3-Max: Just Scale itproblems from Feb to May 2025Source image | CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| Qwen3-VL-235B-A22B-Thinking23 Sep 2025 · Alibaba | 70.1% | Qwen3-VL: Sharper Vision, Deeper Thought, Broader ActionSource image | CheckedA person checked this number against its source on 28 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| Qwen3-Next11 Sep 2025 · Alibaba | 68.7% | Qwen3-Next: Towards Ultimate Training & Inference Efficiencythinking model; LiveCodeBench v6, problems from Feb to May 2025Source image | CheckedA person checked this number against its source on 28 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.CorroboratedThe same number is printed in 1 other lab document: 1 |
| Qwen3-Next11 Sep 2025 · Alibaba | 56.6% | Qwen3-Next: Towards Ultimate Training & Inference Efficiencyinstruct model; LiveCodeBench v6, problems from Feb to May 2025; another report, not chartedSource image | CheckedA person checked this number against its source on 28 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| GLM-4.528 Jul 2025 · Z.ai | 63.3% | GLM-4.6: Advanced Agentic, Reasoning and Coding Capabilities128K context; reported laterSource image | CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| Qwen3-235B-A22B-Thinking-250725 Jul 2025 · Alibaba | 74.1% | Alibaba Unveils New Qwen3 Models for Coding, Complexing Reasoning and Machine Translationproblems from Feb to May 2025Source image | CorroboratedThe same number is printed in 1 other lab document: 1 |
| Kimi K211 Jul 2025 · Moonshot AI | 53.7% | Kimi K2: Open Agentic Intelligenceproblems from Aug 2024 to May 2025; non-thinkingOur capture | |
| Magistral Medium10 Jun 2025 · Mistral | 50.3% | Magistralaveraged over 16 runs | |
| Magistral Small10 Jun 2025 · Mistral | 47.4% | Magistral | |
| Mistral Medium 37 May 2025 · Mistral | 30.3% | Medium is the new large.0-shotSource image | CheckedA person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| Gemini 2.5 Pro25 Mar 2025 · Google | 74.2% | Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilitiespass@1; single attempt; problems from 1/1/2025 to 5/1/2025; reported later | CorroboratedThe same number is printed in 1 other lab document: 1 |
| Gemma 3 27B12 Mar 2025 · Google | 29.1% | Gemma 4: Byte for byte, the most capable open modelsreported laterSource image |
