Humanity's Last Exam
Created by Center for AI Safety and Scale AI, 23 Jan 2025 · Higher is better · Website
- OpenAI
- Other labs
- Reported in a later post or card
Results
| Model | Score | Source |
|---|---|---|
| Qwen3.8-Max3 Aug 2026 · Alibaba | 43.6% | Qwen3.8-Max: A New Bar for Coding and CoworkChecked A person checked this number against its source on 17 Oct 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.full or text-only set not statedOur capture |
| DeepSeek-V4.1-Flash-101717 Oct 2026 · DeepSeek | 42.6% | DeepSeek-V4.1-Flash-1017: Wish-Driven Programming Made RealCorroborated The same number is printed in 1 other lab document: 1Source image |
| Qwen3.7-Max20 May 2026 · Alibaba | 41.4% | Qwen3.7: The Agent FrontierChecked 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.xhigh reasoning effortOur capture |
| GPT-5.523 Apr 2026 · OpenAI | 41.4% | Introducing GPT-5.5Checked A person checked this number against its source on 17 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 6 other lab documents. The first 3 are linked: 1 2 3Our capture |
| GPT-5.45 Mar 2026 · OpenAI | 39.8% | Introducing GPT-5.4Checked A person checked this number against its source on 16 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 5 other lab documents. The first 3 are linked: 1 2 3Our capture |
| DeepSeek-V4-Pro24 Apr 2026 · DeepSeek | 37.7% | DeepSeek V4 Preview ReleaseCorroborated The same number is printed in 1 other lab document: 1max reasoning effort; full or text-only set not statedSource image |
| DeepSeek-V4.1-Flash10 Sep 2026 · DeepSeek | 36.8% | DeepSeek-V4.1-Flash: Smarter, Faster, More EfficientCorroborated The same number is printed in 3 other lab documents: 1 2 3Source image |
| Qwen3.8-Omni-Flash18 Sep 2026 · Alibaba | 36.5% | Qwen3.8-Omni-Flash: Omni Senses. Agentic Delivery.Checked 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.GPT-4o judge; full or text-only set not statedOur capture |
| Qwen3-Max-Thinking26 Jan 2026 · Alibaba | 36.5% | Pushing Qwen3-Max-Thinking Beyond its LimitsChecked 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.heavy mode: test-time scaling over parallel trajectories; parallel test-time computeSource image |
| Qwen3.8-Flash-Next26 Aug 2026 · Alibaba | 35.9% | Qwen3.8-Flash-Next: A New Architecture, Towards Ultimate Cost-EfficiencyCorroborated The same number is printed in 1 other lab document: 1GPT-4o judge; full or text-only set not statedOur capture |
| Qwen3.7-Plus1 Jun 2026 · Alibaba | 34.7% | Qwen3.7-Plus: Multimodal Agent IntelligenceChecked 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: 1xhigh reasoning effort; full or text-only set not statedOur capture |
| GPT-5.211 Dec 2025 · OpenAI | 34.5% | Introducing GPT-5.2Checked A person checked this number against its source on 16 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 3 other lab documents: 1 2 3Our capture |
| Qwen3.6-Plus2 Apr 2026 · Alibaba | 28.8% | Qwen3.6-Plus: Towards Real World AgentsChecked 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.Our capture |
| Qwen3.5-397B-A17B16 Feb 2026 · Alibaba | 28.7% | Qwen3.5: Towards Native Multimodal AgentsChecked 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.Our capture |
| GPT-5.4 mini17 Mar 2026 · OpenAI | 28.2% | Introducing GPT-5.4 mini and nanoChecked 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.xhigh effortOur capture |
| Nemotron 3 Ultra 550B-A55B4 Jun 2026 · NVIDIA | 26.7% | nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4 model card (launch-day revision)Corroborated The same number is printed in 2 other lab documents: 1 2Our capture |
| GPT-5.112 Nov 2025 · OpenAI | 25.7% | Introducing GPT-5.2Checked A person checked this number against its source on 17 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: 1reported laterOur capture |
| GLM-4.722 Dec 2025 · Z.ai | 24.8% | GLM-4.7: Advancing the Coding CapabilityChecked 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: 1Our capture |
| GPT-5.4 nano17 Mar 2026 · OpenAI | 24.3% | Introducing GPT-5.4 mini and nanoChecked A person checked this number against its source on 26 Oct 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.xhigh effortOur capture |
| MiniMax M2.123 Dec 2025 · MiniMax | 22.2% | MiniMax M2.5: Built for Real-World Productivity.Corroborated The same number is printed in 1 other lab document: 1in-house run with Artificial Analysis Intelligence Index methods; reported laterSource image |
| DeepSeek-V3.1-Terminus22 Sep 2025 · DeepSeek | 21.7% | DeepSeek-V3.1-TerminusCorroborated The same number is printed in 2 other lab documents: 1 2reasoning mode; full or text-only set not statedSource image |
| Qwen3.6-35B-A3B15 Apr 2026 · Alibaba | 21.4% | Qwen3.6-35B-A3B: Agentic Coding Power, Now Open to AllChecked 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: 1full or text-only set not statedOur capture |
| OpenAI o316 Apr 2025 · OpenAI | 20.32% | Introducing OpenAI o3 and o4-miniCorroborated The same number is printed in 3 other lab documents: 1 2 3high reasoning effortOur capture |
| OpenAI o316 Apr 2025 · OpenAI | 20.2% | Introducing GPT-5 for developersChecked 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.high reasoning effort; reported later; another report, not chartedOur capture |
| DeepSeek-V3.2-Exp29 Sep 2025 · DeepSeek | 19.8% | Introducing DeepSeek-V3.2-ExpCorroborated The same number is printed in 1 other lab document: 1full or text-only set not statedSource image |
| MiniMax M2.512 Feb 2026 · MiniMax | 19.4% | MiniMax M2.5: Built for Real-World Productivity.Checked 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.in-house run with Artificial Analysis Intelligence Index methods; full or text-only set not statedSource image |
| Gemini 2.5 Pro25 Mar 2025 · Google | 18.8% | Gemini 2.5: Our most intelligent AI modelChecked A person checked this number against its source on 17 Oct 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.pass@1Source image |
| Nemotron 3 Super 120B-A12B11 Mar 2026 · NVIDIA | 18.26% | nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8 model card (launch-day revision)Corroborated The same number is printed in 3 other lab documents: 1 2 3Our capture |
| GLM-4.630 Sep 2025 · Z.ai | 17.2% | GLM-4.6: Advanced Agentic, Reasoning and Coding CapabilitiesChecked 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: 1128K contextSource image |
| GPT-5 mini7 Aug 2025 · OpenAI | 16.7% | Introducing GPT-5 for developersCorroborated The same number is printed in 1 other lab document: 1high reasoning effort; reported laterOur capture |
| GPT-5 mini7 Aug 2025 · OpenAI | 18.3% | Introducing GPT-5.4 mini and nanoChecked 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.high reasoning effort, the highest available; reported later; another report, not chartedOur capture |
| gpt-oss-120b5 Aug 2025 · OpenAI | 14.9% | Introducing gpt-ossChecked 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.Our capture |
| o4-mini16 Apr 2025 · OpenAI | 14.28% | Introducing OpenAI o3 and o4-miniCorroborated The same number is printed in 2 other lab documents: 1 2high reasoning effortOur capture |
| Qwen3-VL-235B-A22B-Thinking23 Sep 2025 · Alibaba | 13.6% | Qwen3-VL: Sharper Vision, Deeper Thought, Broader ActionChecked 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.thinking model; full or text-only set not statedSource image |
| o3-mini31 Jan 2025 · OpenAI | 13.40% | Introducing OpenAI o3 and o4-miniCorroborated The same number is printed in 1 other lab document: 1high reasoning effort; reported laterOur capture |
| Gemini 2.5 Flash17 Apr 2025 · Google | 12.1% | Developers can now start building with Gemini 2.5 Flash.Corroborated The same number is printed in 1 other lab document: 1thinking; pass@1Source image |
| Nemotron 3 Nano 30B-A3B15 Dec 2025 · NVIDIA | 10.2% | nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-FP8 model card (launch-day revision)Corroborated The same number is printed in 1 other lab document: 1Our capture |
| Nemotron 3 Nano 30B-A3B15 Dec 2025 · NVIDIA | 10.57% | Nemotron 3 Nano: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Model for Agentic ReasoningCorroborated The same number is printed in 1 other lab document: 1another report, not chartedOur capture |
| GPT-5 nano7 Aug 2025 · OpenAI | 8.7% | Introducing GPT-5 for developersChecked 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.high reasoning effort; reported laterOur capture |
| Gemini 2.5 Flash-Lite17 Jun 2025 · Google | 6.9% | We’re expanding our Gemini 2.5 family of modelsChecked A person checked this number against its source on 26 Oct 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.thinking; pass@1Source image |
| Gemini 2.5 Flash-Lite17 Jun 2025 · Google | 5.1% | We’re expanding our Gemini 2.5 family of modelsChecked A person checked this number against its source on 26 Oct 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.non-thinking; pass@1; another report, not chartedSource image |
