Humanity's Last Exam
Created by Center for AI Safety and Scale AI, 23 Jan 2025 · Higher is better · Website
- Other labs
- Reported in a later post or card
Results
| Model | Score | Source |
|---|---|---|
| LongCat-2.51 Nov 2026 · Meituan | 63.1% | LongCat-2.5: Native Multimodality for Agentic TasksChecked A person checked this number against its source on 2 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.text-only subset; Claude Code harness; 256K output budgetOur capture |
| Hy4 preview28 Aug 2026 · Tencent | 55.4% | Introducing Hy4 previewChecked A person checked this number against its source on 9 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.text-only subset; highest reasoning settingSource image |
| Hy4 preview28 Aug 2026 · Tencent | 57.3% | Introducing Hy4Checked A person checked this number against its source on 8 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.re-evaluated at Hy4 launch; text-only subset; reported later; another report, not chartedSource image |
| GLM-5.216 Jun 2026 · Z.ai | 54.7% | GLM-5.2: Built for Long-Horizon TasksChecked A person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Corroborated The same number is printed in 2 other lab documents: 1 2text-only subset; 300,000-token context, no context managementOur capture |
| Hy36 Jul 2026 · Tencent | 53.2% | Introducing Hy3Corroborated The same number is printed in 1 other lab document: 1text-only subset; highest reasoning settingSource image |
| Hy36 Jul 2026 · Tencent | 35.4% | Introducing Hy3Corroborated The same number is printed in 1 other lab document: 1Hy3 preview, re-evaluated at Hy3 launch; text-only subset; highest reasoning setting; another report, not chartedSource image |
| Hy36 Jul 2026 · Tencent | 51.9% | Introducing Hy4 previewChecked A person checked this number against its source on 7 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.re-evaluated at Hy4 preview launch; text-only subset; highest reasoning setting; reported later; another report, not chartedSource image |
| GLM-5.17 Apr 2026 · Z.ai | 52.3% | GLM-5.1: Towards Long-Horizon TasksChecked A person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Corroborated The same number is printed in 1 other lab document: 1text-only subset; GPT-5.2 judge; 202,752-token contextOur capture |
| GLM-512 Feb 2026 · Z.ai | 50.4% | GLM-5: From Vibe Coding to Agentic EngineeringChecked A person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Corroborated The same number is printed in 1 other lab document: 1text-only subset; GPT-5.2 judge; 202,752-token contextOur capture |
| Nemotron 3.5 Super VL 120B-A12B15 Oct 2026 · NVIDIA | 46.01% | Nemotron 3.5 Super and Lightning: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Models for Agentic AIChecked A person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.text-only subset; Python and web search; separate internal harnessOur capture |
| Kimi K2 Thinking6 Nov 2025 · Moonshot AI | 44.9% | Introducing Kimi K2 ThinkingCorroborated The same number is printed in 1 other lab document: 1text-only subset; search, code and browsing tools; up to 120 steps; Hugging Face blocked; INT4 precisionOur capture |
| Kimi K2 Thinking6 Nov 2025 · Moonshot AI | 51.3% | Introducing Kimi K2 ThinkingChecked A person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Hugging Face access not blocked, which may leak answers; text-only subset; another report, not charted |
| Kimi K2 Thinking6 Nov 2025 · Moonshot AI | 51.0% | Introducing Kimi K2 ThinkingChecked A person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.heavy mode: eight parallel trajectories aggregated into one answer; text-only subset; parallel test-time computeOur capture |
| DeepSeek-V3.121 Aug 2025 · DeepSeek | 29.8% | DeepSeek-V3.1 ReleaseChecked A person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.search agentSource image |
| DeepSeek-R1-052828 May 2025 · DeepSeek | 24.8% | DeepSeek-V3.1 ReleaseCorroborated The same number is printed in 1 other lab document: 1search agent; reported laterSource image |
