Humanity's Last Exam
last retrieved 2026-09-08 21:40 UTC
Third-party scores shown as context with provenance — our own arena ratings from blind human votes stay alongside every row and remain the primary signal.
| Model | Humanity's Last Exam | Observed | Source |
|---|---|---|---|
| Claude Fable 5.1Anthropic | 46.5%42.6% – 50.4% | 2026-09-08 | source ↗ |
| Gemini 3.1 Pro PreviewGoogle | 46.4%42.6% – 50.3% | 2026-09-08 | source ↗ |
| GPT-5.4OpenAI | 36.2%32.6% – 39.9% | 2026-09-08 | source ↗ |
| Claude Opus 4.7Anthropic | 36.2%32.5% – 39.9% | 2026-09-08 | source ↗ |
| Llama 4 MaverickMeta | 5.7%3.9% – 7.5% | 2026-09-08 | source ↗ |
— = not yet evaluated by HumanEval (no eligible arena votes). External values are the newest observation per model; hover a value for source notes and a rating for its 95% CI. Unit: %, higher is better.
Not enough overlap with our arena ratings for a correlation scatter yet — fewer than two scored models are rated.