ArXivLean
Models prove or disprove theorems from recent arXiv papers, stated in Lean 4. A strict comparator checks each submitted proof against the fixed reference statement.
The first set held 41 theorems from March 2026 papers, and the best systems proved about 17 percent. Later sets came in June and September 2026.
- Level
- Research frontier Researchers, on open problems or tasks that take days
- Scored
- Checked automatically A program checks the result against an answer key, unit tests or the expected final state.
- Answers
- Written proof
Claude Opus 5.5 leads with 47.2%
Best score per model from the benchmark's own leaderboard, read on 2 Nov 2026. Where a model ran with several settings or harnesses, we keep its best.
- 1 Claude Opus 5.5 high 47.2%
- 2
Gemini 4 Argon 39.6% - 3 GPT-6 Astra max 37.7%
- 4 GPT-6.1 Sol max 35.9%
- 5 Kimi K3.1 Pro 17%
- 6
DeepSeek V4.1 Flash max 5.66% - 7
Qwen3.8-2.4T-A95B 1.89%
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude Opus 5.5 | 47.2% | high | Sep 2026 | Official leaderboard |
| 2 | Gemini 4 Argon | 39.6% | Sep 2026 | Official leaderboard | |
| 3 | GPT-6 Astra | 37.7% | max | Sep 2026 | Official leaderboard |
| 4 | GPT-6.1 Sol | 35.9% | max | Sep 2026 | Official leaderboard |
| 5 | Kimi K3.1 Pro | 17% | Oct 2026 | Official leaderboard | |
| 6 | DeepSeek V4.1 Flash | 5.66% | max | Sep 2026 | Official leaderboard |
| 7 | Qwen3.8-2.4T-A95B | 1.89% | Aug 2026 | Official leaderboard |
Labs report scores for 1 model in their cards. We plot every source in the charts below.
3 models have held number one since its release in Apr 2026
Lab claims count from the day the card came out, independent runs from the model’s release.
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Labs report it only in system cards
We read the headline results of 479 launch posts from 51 labs, and none of them list this benchmark. Some system cards do, below.
One system card reports it
Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| 2 Jul 2026 |  Leanstral 1.5 technical report | 17.1 | ArXivLean, 7 of 41 problems (3.1.4) | Capabilities |
More in Math
- IMO-AnswerBench In use 400 olympiad problems with short verifiable answers, robustified against memorization. Scored by answer match. GLM-5.2 91%
- AIME 2026 In use Models solve the 30 integer-answer problems from the 2026 AIME I and II math competitions. Scored by exact answer match. GPT-5.5 100%
- HMMT 2026 In use Models solve the 33 final-answer problems from the February 2026 Harvard-MIT Mathematics Tournament. Scored by exact answer match. GPT-5.5 98.5%
- MathArena (Apex, IMO, USAMO) In use Evaluates models on newly released math competitions after release. Final-answer or human/LLM-judged proof scores. Claude Opus 4.8 81.3%
- PolyMath In use Models solve 9,000 parallel math problems in 18 languages across four difficulty levels, from school math to frontier math. Final answers are checked. Qwen3.7-Max 86.5%
- USAMO 2026 In use Models write full proofs for the six problems of the 2026 USA Mathematical Olympiad. Judges grade each proof from 0 to 7 against rubrics. GPT-5.5 98.2%