Humanity's Last Exam leaderboard
As of November 8, 2026, Kimi K3.1 Pro (Moonshot) leads Humanity's Last Exam at 70%, ahead of Claude Opus 5.5 at 67.7% and Claude Fable 5.1 at 65.6%. 20 systems ranked. Every score here is published by the model's own lab; each row links to the page it comes from.
What is Humanity's Last Exam? Humanity's Last Exam is a set of 2,500 expert-written questions across dozens of academic subjects, built to be hard for frontier models; each has one checkable answer. Made by Center for AI Safety and Scale AI, with expert contributors worldwide, January 2025.
Top models on Humanity's Last Exam, November 2026
The model may search the web or run code.
| # | Model | Score | Price / 1M | Context | Source |
|---|---|---|---|---|---|
| 1 | Kimi K3.1 Pro Moonshot (Kimi) | max effort, with tools | $2 / $10 | 1M | Vendor-reported Moonshot AI |
| 2 | Claude Opus 5.5 Anthropic | with tools, max effort | $4 / $20 | 1M | Vendor-reported Anthropic |
| 3 | Claude Fable 5.1 Anthropic | with tools, max effort | $10 / $50 | 1M | Vendor-reported Anthropic |
| 4 | Claude Sonnet 5.5 Anthropic | with tools | $2 / $10 | 1M | Vendor-reported Anthropic |
| 5 | DeepSeek V4.1 Flash DeepSeek | pass@1, with tools | $0.3 / $1.2 | 1M | Vendor-reported DeepSeek |
| 6 | Claude Opus 5 Anthropic | with tools, max effort | $5 / $25 | 1M | Vendor-reported Anthropic |
| 7 | GLM-5.3 Zhipu (Z.ai / GLM) | with tools | $1.4 / $4.4 | 1M | Vendor-reported Z.ai (Zhipu) |
| 8 | DeepSeek V4 Pro DeepSeek | pass@1, with tools | $1.32 / $3.96 | 1M | Vendor-reported DeepSeek |
| 9 | Claude Haiku 5.5 Anthropic | with tools | $0.1 / $0.5 | 1M | Vendor-reported Anthropic |
| 10 | GPT-6 Astra OpenAI | with tools | $10 / $50 | 1.05M | Vendor-reported OpenAI |
| 11 | Qwen3.8 Max Alibaba (Qwen) | with tools | $2 / $6 | 1M | Vendor-reported Alibaba (Qwen) |
| 12 | Kimi K3 Moonshot (Kimi) | max effort, with tools | $3 / $15 | 1M | Vendor-reported Moonshot AI |
| 13 | Claude Sonnet 5 Anthropic | with tools | $2 / $10 | 1M | Vendor-reported Anthropic |
| 14 | GLM-5.2 Zhipu (Z.ai / GLM) | with tools | $1.4 / $4.4 | 1M | Vendor-reported Z.ai (Zhipu) |
| 15 | Qwen3.7 Max Alibaba (Qwen) | with tools | $2.5 / $7.5 | 1M | Vendor-reported Alibaba (Qwen) |
| 16 | DeepSeek V4 Flash Deprecated DeepSeek | pass@1, with tools | $0.44 / $1.32 | 1M | Vendor-reported DeepSeek |
| 17 | Gemini 3.1 Pro Preview Preview Google · listed as “Gemini 3.1 Pro” | search and code, thinking high | $2 / $12 | 1M | Vendor-reported Google DeepMind |
| 18 | Gemini 3 Pro Preview Retired Google · listed as “Gemini 3 Pro” | search and code, thinking high | $2 / $12 | 1M | Vendor-reported Google DeepMind |
| 19 | Claude Haiku 4.5 Anthropic | with tools | $1 / $5 | 200K | Vendor-reported Anthropic |
Humanity's Last Exam under other settings
Labs report Humanity's Last Exam under different settings, and a score from one setting is not comparable with a score from another. Each setting has its own table, ranked on its own.
No tools
| # | Model | Score | Price / 1M | Context | Source |
|---|---|---|---|---|---|
| 1 | Claude Sonnet 5.5 Anthropic | no tools | $2 / $10 | 1M | Vendor-reported Anthropic |
| 2 | Kimi K3.1 Pro Moonshot (Kimi) | max effort, no tools | $2 / $10 | 1M | Vendor-reported Moonshot AI |
| 3 | Claude Haiku 5.5 Anthropic | no tools | $0.1 / $0.5 | 1M | Vendor-reported Anthropic |
| 4 | Gemini 3.1 Pro Preview Preview Google · listed as “Gemini 3.1 Pro” | no tools, thinking high | $2 / $12 | 1M | Vendor-reported Google DeepMind |
| 5 | Qwen3.8 Max Alibaba (Qwen) | no tools | $2 / $6 | 1M | Vendor-reported Alibaba (Qwen) |
| 6 | Kimi K3 Moonshot (Kimi) | max effort, no tools | $3 / $15 | 1M | Vendor-reported Moonshot AI |
| 7 | MiniMax M3.1 Flash MiniMax | no tools | $0.15 / $0.6 | 1M | Vendor-reported MiniMax |
| 8 | DeepSeek V4.1 Flash DeepSeek · listed as “DeepSeek V4.1 Flash (1017)” | no tools | $0.3 / $1.2 | 1M | Vendor-reported DeepSeek |
| 9 | Qwen3.7 Max Alibaba (Qwen) | no tools | $2.5 / $7.5 | 1M | Vendor-reported Alibaba (Qwen) |
| 10 | Gemini 3 Pro Preview Retired Google · listed as “Gemini 3 Pro” | no tools, thinking high | $2 / $12 | 1M | Vendor-reported Google DeepMind |
| 11 | Claude Haiku 4.5 Anthropic | no tools | $1 / $5 | 200K | Vendor-reported Anthropic |
Humanity's Last Exam: score against price
Can you trust Humanity's Last Exam scores?
No independent leaderboard publishes Humanity's Last Exam results for these models yet, so every number on this page is the lab's own report of its own model. Labs choose the harness, the effort level and the subset they report, and they rarely publish failures. Read the setting under each score before comparing two rows.
- Labs report HLE with tools (search, code) and without, and the two are far apart. The official site does not define either setting, so they are separate tables here.
- Models that cannot read images are scored on the text-only subset, so not every score uses the same questions.
- The authors keep a private held-out set to check for overfitting to the public questions.
How Humanity's Last Exam works
The task
A closed-ended question, multiple choice or exact short answer, in mathematics, the sciences or the humanities; some include an image. The model has to give the single correct answer.
The score
Accuracy, graded automatically by a judge model against the known answer. The official site also reports calibration error from the model's stated confidence.
Test set
2,500 questions, as published by the authors.
Made by
Center for AI Safety and Scale AI, with expert contributors worldwide, first published January 2025.
Official sources
- lastexam.ai
- Paper: arxiv.org/abs/2501.14249
- Definition read on 2026-11-08
Sources and freshness
31 published scores from 12 sources. Each was read on its source page; the oldest reading on this page is from November 8, 2026. Prices beside each score come from our model price index and move when a vendor reprices.
- Vendor-reported Anthropic
- Vendor-reported Alibaba (Qwen)
- Vendor-reported DeepSeek
- Vendor-reported Google DeepMind
- Vendor-reported Anthropic
- Vendor-reported Anthropic
- Vendor-reported Moonshot AI
- Vendor-reported Moonshot AI
- Vendor-reported Z.ai (Zhipu)
- Vendor-reported DeepSeek
- Vendor-reported MiniMax
- Vendor-reported OpenAI
Humanity's Last Exam FAQ
Answers derived from the 31 published scores on this page and the current price index.