Humanity's Last Exam (Scale run) leaderboard
- GPT-6 Astra leads
- 60.6%
- models measured
- 16
- lead over #2
- +5.6 pts
- median
- 32.5%
Leader on Humanity's Last Exam (Scale run) · read October 22
On October 22, 2026, 16 of our catalog's models had a Humanity's Last Exam (Scale run) result, and OpenAI GPT-6 Astra led with 60.6%, ahead of Anthropic Claude Opus 5.5 at 55.0%.
Humanity's Last Exam asks expert-level questions across science, math and the humanities. This is Scale AI's own run of it. The number is the share answered correctly.
Role in the index
Humanity's Last Exam (Scale run) is shown as published and does not feed any of our indexes. Read the methodology
| # | Model | Result | Evidence | Coding index | Output $/Mtok |
|---|---|---|---|---|---|
| 1 | GPT-6 AstraOpenAI | 60.6% | 85.1 | $50.00 | |
| 2 | Claude Opus 5.5Anthropic | 55.0% | 88.8 | $20.00 | |
| 3 | Gemini 4 ArgonGoogle | 53.2% | 81.0 | $10.00 | |
| 4 | Claude Fable 5.1Anthropic | 51.3% | 84.2 | $50.00 | |
| 5 | Claude Opus 5Anthropic | 38.6% | 83.0 | $25.00 | |
| 6 | Claude Sonnet 5.5Anthropic | 37.5% | 87.3 | $10.00 | |
| 7 | Gemini 3.8 FlashGoogle | 34.3% | 61.6 | $3.75 | |
| 8 | GPT-6 SolOpenAI | 33.8% | 76.0 | $10.00 | |
| 9 | GPT-5.6 SolOpenAI | 31.2% | 73.5 | $20.00 | |
| 10 | Muse Spark OpenMeta | 28.0% | 47.4 | $4.25 | |
| 11 | Muse Spark 1.3Meta | 25.4% | 57.4 | $4.25 | |
| 12 | Grok 4.7xAI | 23.4% | 66.7 | $6.00 | |
| 13 | Kimi K3Moonshot AI | 22.2% | 62.3 | $15.00 | |
| 14 | GLM 5.3Z.ai | 16.4% | 70.6 | $4.40 | |
| 15 | DeepSeek V4 Pro 0423DeepSeek | 13.4% | 38.7 | $2.60 | |
| 16 | T InklingThinking Machines | 11.2% | 31.5 | $4.05 |
Only models in our catalog are listed, and each appears once with its best published run. 66 rows were read from the source on October 22, 2026, and rows that match no catalog model are left out.