Benchmarks / Knowledge & Science
Humanity's Last Exam
Frontier-difficulty closed-ended academic questions (CAIS × Scale).
maintainer unknown · 284 observations on record
Versions
1 version
| Version | Tasks |
|---|---|
| Humanity's Last Exam current humanitys-last-exam | — |
Current leaderboard
Humanity's Last Exam · Accuracy · top 25 of 276 systems
| # | System | Accuracy % |
|---|---|---|
| 1 | GPT 5.6 Solmax effort OpenAI | 47.2% |
| 2 | Gemini 3.1 Pro Previewhigh effort Google DeepMind | 46.4% 44.5–48.4 |
| 3 | Claude Opus 4.8max effort Anthropic | 45.7% |
| 4 | Muse Spark 1.1xhigh effort Meta AI | 45.1% |
| 5 | Gemini 3.1 Pro Preview Google DeepMind | 44.7% |
| 6 | GPT 5.6 Solxhigh effort OpenAI | 44.7% |
| 7 | GPT 5.4 Pro (2026-03-05) OpenAI | 44.3% 42.4–46.3 |
| 8 | GPT 5.5 (2026-04-22)xhigh effort OpenAI | 44.3% |
| 9 | GPT 5.6 Solhigh effort OpenAI | 44.1% |
| 10 | GPT 5.5 (2026-04-22)high effort OpenAI | 43.0% |
| 11 | GPT 5.6 Terramax effort OpenAI | 41.8% |
| 12 | GPT 5.4 (2026-03-05)xhigh effort OpenAI | 41.6% |
| 13 | Gemini 3.5 Flashhigh effort Google DeepMind | 41.0% |
| 14 | GPT 5.5 (2026-04-22)medium effort OpenAI | 40.6% |
| 15 | Muse Spark Meta AI | 40.6% 38.6–42.5 |
| 16 | Grok 4.5high effort xAI | 40.3% |
| 17 | GLM 5.2max effort Z.ai | 40.1% |
| 18 | GPT 5.6 Terraxhigh effort OpenAI | 40.0% |
| 19 | GPT 5.3 Codexxhigh effort OpenAI | 39.9% |
| 20 | Gemini 3.5 Flashmedium effort Google DeepMind | 39.9% |
| 21 | GPT 5.6 Solmedium effort OpenAI | 39.7% |
| 22 | Claude Sonnet 5max effort Anthropic | 39.6% |
| 23 | Claude Opus 4.7max effort Anthropic | 39.6% |
| 24 | Qwen3 7 Max Alibaba (Qwen) | 38.1% |
| 25 | Gemini 3 Pro Preview Google DeepMind | 37.5% 35.6–39.4 |
One row per evaluated system — reasoning-effort variants rank separately. Where a system has attr variants for this version (eval splits, style control), only the canonical variant is ranked: split=semi_private, style_control=true, or the unsplit run.
Source agreement
where evaluating organizations agree — and don't
Humanity's Last ExamArtificial Analysis vs Scale Labs
overlap 18 systemsrank correlation 0.98mean |Δ| 2.46max |Δ| 7.6
most contested: GPT 5.2 (2025-12-11) (+7.6) · Gemini 3.1 Flash Lite Preview (+7.56) · GPT 5.4 (2026-03-05) (+5.36)
Possible reasons: run setting 'implementation' differs (artificial-analysis vs scale); run setting 'contamination' differs (— vs Potential contamination warning: This model was evaluated after the public release of HLE, allowing model builder access to the prompts and solutions.).