benchmark evidence
Humanity's Last Exam
Frontier multi-domain benchmark. 2,500 expert-crafted questions across math, science, humanities. Published in Nature 2026.
activeupstream source ->
winner on Humanity's Last Exam
direct benchmark result, not a broad vertical composite | source row dated 2000-01-01
scored on 2025-04-03 · stale source data (422d)
latest mapped results | top 20
| # | Model | Score | Evidence | Tested |
|---|---|---|---|---|
| 1 | Anthropic: Claude Opus 4.7 | 54.7 | model-only independent_benchmark | 2000-01-01 |
| 2 | Anthropic: Claude Opus 4.6 | 53.1 | model-only independent_benchmark | 2000-01-01 |
| 3 | OpenAI: GPT-5.5 | 52.2 | model-only independent_benchmark | 2000-01-01 |
| 4 | MoonshotAI: Kimi K2 Thinking | 51.0 | model-only independent_benchmark | 2000-01-01 |
| 5 | Anthropic: Claude Sonnet 4.6 | 49.0 | model-only independent_benchmark | 2000-01-01 |
| 6 | DeepSeek: DeepSeek V4 Pro | 48.2 | model-only independent_benchmark | 2000-01-01 |
| 7 | DeepSeek: DeepSeek V4 Flash | 45.1 | model-only independent_benchmark | 2000-01-01 |
| 8 | DeepSeek: DeepSeek V3.2 | 40.8 | model-only independent_benchmark | 2000-01-01 |
| 9 | OpenAI: GPT-5.4 | 39.8 | model-only independent_benchmark | 2000-01-01 |
| 10 | OpenAI: GPT-5.2 | 34.5 | model-only independent_benchmark | 2000-01-01 |
| 11 | OpenAI: GPT-5.4 Mini | 28.2 | model-only independent_benchmark | 2000-01-01 |
| 12 | OpenAI: GPT-5.1 | 25.3 | model-only independent_benchmark | 2025-04-03 |
| 13 | Google: Gemini 2.5 Pro | 21.6 | model-only independent_benchmark | 2025-04-03 |
| 14 | Anthropic: Claude Sonnet 4.5 | 13.7 | model-only independent_benchmark | 2025-04-03 |
| 15 | Google: Gemini 2.5 Flash | 12.1 | model-only independent_benchmark | 2025-04-03 |
| 16 | DeepSeek: R1 | 8.5 | model-only independent_benchmark | 2025-04-03 |
| 17 | OpenAI: GPT-5 | 2.7 | model-only independent_benchmark | 2025-04-03 |
what this result means
Frontier multi-domain benchmark. 2,500 expert-crafted questions across math, science, humanities. Published in Nature 2026.
This benchmark contributes direct public evidence. Read its scope before generalizing the result.
A win here is a win on Humanity's Last Exam. Broad task pages require independent corroboration before naming a general winner.
source record
category: reasoning
metric: accuracy
matched models: 17
latest source date: 2025-04-03
direction: higher is better