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 |
|---|---|---|
| DeepSeek-V4.1-Flash10 Sep 2026 · DeepSeek | 63.9% | DeepSeek-V4.1-Flash: Smarter, Faster, More EfficientCorroborated The same number is printed in 2 other lab documents: 1 2Source image |
| GLM-5.314 Aug 2026 · Z.ai | 62.5% | GLM-5.3: Frontier Coding with Emergent Cyber 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: 1300,000-token context with context management; GPT-5.6 Luna judgeOur capture |
| DeepSeek-V4-Pro-081313 Aug 2026 · DeepSeek | 60.0% | DeepSeek-V4-Pro GA ReleaseCorroborated The same number is printed in 2 other lab documents: 1 2Source image |
| GPT-6 Astra3 Sep 2026 · OpenAI | 57.2% | GPT-6 Astra: A new generation of intelligenceChecked 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 2 other lab documents: 1 2Our capture |
| Qwen3.8-Max3 Aug 2026 · Alibaba | 56.2% | Qwen3.8-Max: A New Bar for Coding and CoworkChecked 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: 1tools not statedOur capture |
| Qwen3.7-Max20 May 2026 · Alibaba | 53.5% | 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.Corroborated The same number is printed in 1 other lab document: 1Our capture |
| GPT-5.523 Apr 2026 · OpenAI | 52.2% | 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 | 52.1% | Introducing GPT-5.4Checked 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 5 other lab documents. The first 3 are linked: 1 2 3Our capture |
| Qwen3.6-Plus2 Apr 2026 · Alibaba | 50.6% | 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.Corroborated The same number is printed in 1 other lab document: 1256K context with context foldingOur capture |
| Qwen3.6-Plus2 Apr 2026 · Alibaba | 50.2% | 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.reported later; another report, not chartedOur capture |
| Qwen3-Max-Thinking26 Jan 2026 · Alibaba | 49.8% | 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.with search, HLE-related websites blockedOur capture |
| Qwen3-Max-Thinking26 Jan 2026 · Alibaba | 58.3% | 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; with search; parallel test-time computeSource image |
| Qwen3.5-397B-A17B16 Feb 2026 · Alibaba | 48.3% | 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.search agent; context folding at 256K tokensOur capture |
| DeepSeek-V4-Pro24 Apr 2026 · DeepSeek | 48.2% | DeepSeek V4 Preview ReleaseCorroborated The same number is printed in 1 other lab document: 1max reasoning effortSource image |
| Inkling15 Jul 2026 · Thinking Machines | 46.0% | Inkling: Our Open-Weights ModelCorroborated The same number is printed in 2 other lab documents: 1 2effort 0.99; tools not statedOur capture |
| GPT-5.211 Dec 2025 · OpenAI | 45.5% | 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 7 other lab documents. The first 3 are linked: 1 2 3search and PythonOur capture |
| GLM-4.722 Dec 2025 · Z.ai | 42.8% | 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-5.112 Nov 2025 · OpenAI | 42.7% | 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: 1search and Python; reported laterOur capture |
| GPT-5.4 mini17 Mar 2026 · OpenAI | 41.5% | 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 effort; tools not namedOur capture |
| GPT-5.4 nano17 Mar 2026 · OpenAI | 37.7% | 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 effort; tools not namedOur capture |
| Nemotron 3 Ultra 550B-A55B4 Jun 2026 · NVIDIA | 37.4% | Nemotron 3 Ultra: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Model for Agentic ReasoningCorroborated The same number is printed in 2 other lab documents: 1 2Our capture |
| GPT-5 mini7 Aug 2025 · OpenAI | 31.6% | 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; tools not named; reported laterOur capture |
| GLM-4.630 Sep 2025 · Z.ai | 30.4% | 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 |
| OpenAI o316 Apr 2025 · OpenAI | 24.90% | Introducing OpenAI o3 and o4-miniCorroborated The same number is printed in 2 other lab documents: 1 2high reasoning effort; Python and browsing; known cheating domains blocked; attempts a monitor flags count as wrongOur capture |
| Nemotron 3 Super 120B-A12B11 Mar 2026 · NVIDIA | 22.8% | 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 3Source image |
| gpt-oss-120b5 Aug 2025 · OpenAI | 19% | 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.tools not namedOur capture |
| o4-mini16 Apr 2025 · OpenAI | 17.70% | Introducing OpenAI o3 and o4-miniChecked 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.high reasoning effort; Python and browsing; known cheating domains blocked; attempts a monitor flags count as wrongOur capture |
| Nemotron 3 Nano 30B-A3B15 Dec 2025 · NVIDIA | 15.5% | nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-FP8 model card (launch-day revision)Corroborated The same number is printed in 2 other lab documents: 1 2Our capture |
