Benchmark Almanac

IMO-ProofBench

Aletheia 91.9% tops its leaderboard GPT-5.5 Pro 88.1% leads the models we track
GPT-5.5 Pro 88.1% leads since April 2026
Fading

Models write proofs for 60 olympiad problems, split into a basic set and an advanced set up to IMO-hard. Each proof is graded from 0 to 7 by experts or a Gemini autograder.

Part of IMO-Bench (EMNLP 2025). The v2 file only fixed a typo.

Released
November 2025
Kind
Written proofs
Level
Competition
Scored
Graded by people
Answers
Written proof
Tasks
All public
Strongest model tested
GPT-5.5 Pro #14 on the capability index
Last in a launch post
September 2026, dropped after it saturated
Last model added
September 2026
Scores refreshed
November 12th, 2026

In the authors’ words

Finding the right north-star metrics is highly critical for advancing the mathematical reasoning capabilities of foundation models, especially given that existing evaluations are either too easy or only focus on getting correct short answers. To address these issues, we present IMO-Bench, a suite of advanced reasoning benchmarks, vetted by a panel of top specialists and that specifically targets the level of the International Mathematical Olympiad (IMO), the most prestigious venue for young mathematicians. IMO-AnswerBench first tests models on 400 diverse Olympiad problems with verifiable short answers. IMO-Proof Bench is the next-level evaluation for proof-writing capabilities, which includes both basic and advanced IMO level problems as well as detailed grading guidelines to facilitate automatic grading. These benchmarks played a crucial role in our historic achievement of the gold-level performance at IMO 2025 with Gemini Deep Think (Luong and Lockhart, 2025). Our model achieved 80.0% on IMO-AnswerBench and 65.7% on the advanced IMO-Proof Bench, surpassing the best non-Gemini models by large margins of 6.9% and 42.4% respectively. We also showed that autograders built with Gemini reasoning correlate well with human evaluations and construct IMO-GradingBench, with 1000 human gradings on proofs, to enable further progress in automatic evaluation of long-form answers. We hope that IMO-Bench will help the community towards advancing robust mathematical reasoning and release it at https://imobench.github.io/.

Abstract of Towards Robust Mathematical Reasoning (arXiv:2511.01846, CC0)
26
Entries on its own leaderboard, 20 models we track
3
Launch posts list it, from 3 labs
4
System cards report it, from 3 labs

Aletheia tops the leaderboard with 91.9%

Best score per entry from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. The best model we track is GPT-5.5 Pro, at 88.1%.

GPT-5.5 Pro leads with 88.1%, claims included

Best score per model from every source, including labs’ own numbers and claims Epoch AI collected from reports and papers. Labs pick their own settings, so a claim can beat an independent run of the same model.

The newest model with a score here came out in April 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

26 entries. Scroll sideways for the rest.

  1. 91.9 Aletheia
  2. 88.1 GPT-5.5 Pro
  3. 76.7 Gemini 3 Deep Think
  4. 71.9 GPT-5.5
  5. 65.7 Gemini Deep Think (IMO Gold)
  6. 49.0 Gemini 3.1 Pro
  7. 37.6 Gemini Deep Think (IMO lite)
  8. 35.7 GPT-5.2
  9. 30.0 Gemini 3 Pro
  10. 28.6 GPT-5 Pro
  11. 24.8 Gemini 2.5 Pro with ( Huang & Yang, 2025 )
  12. 23.8 Claude Opus 4.5
  13. 23.3 Grok 4 Heavy
  14. 20.5 o3
  15. 20.0 GPT-5
  16. 18.6 Grok 4
  17. 18.6 Grok 4.1 Fast
  18. 17.6 Gemini 2.5 Pro
  19. 11.4 o4-mini
  20. 7.10 Kimi-K2-Instruct
  21. 7.10 GPT-5.1
  22. 5.20 Qwen3-235B
  23. 4.80 Claude Sonnet 4
  24. 4.30 DeepSeek-V3
  25. 3.80 DeepSeek-R1
  26. 2.90 Claude Opus 4
  1. 1 Aletheia 91.9%
  2. 2 GPT-5.5 Pro xhigh 88.1%
  3. 3 Gemini 3 Deep Think 76.7%
  4. 4 GPT-5.5 xhigh 71.9%
  5. 5 Gemini Deep Think (IMO Gold) 65.7%
  6. 6 Gemini 3.1 Pro 49%
  7. 7 Gemini Deep Think (IMO lite) 37.6%
  8. 8 GPT-5.2 high 35.7%
  9. 9 Gemini 3 Pro 30%
  10. 10 GPT-5 Pro 28.6%
  11. 11 Gemini 2.5 Pro with ( Huang & Yang, 2025 ) 24.8%
  12. 12 Claude Opus 4.5 23.8%
  13. 13 Grok 4 Heavy 23.3%
  14. 14 o3 20.5%
  15. 15 GPT-5 20%
  16. 16 Grok 4 18.6%
  17. 17 Grok 4.1 Fast reasoning 18.6%
  18. 18 Gemini 2.5 Pro Jun 2025 17.6%
  19. 19 o4-mini high 11.4%
  20. 20 Kimi-K2-Instruct 7.10%
  21. 21 GPT-5.1 7.10%
  22. 22 Qwen3-235B 5.20%
  23. 23 Claude Sonnet 4 4.80%
  24. 24 DeepSeek-V3 Mar 2025 4.30%
  25. 25 DeepSeek-R1 May 2025 3.80%
  26. 26 Claude Opus 4 2.90%

Source: Official IMO-ProofBench leaderboard

25 models. Scroll sideways for the rest.

  1. 88.1 GPT-5.5 Pro
  2. 82.3 Nemotron 3 Ultra
  3. 76.7 Gemini 3 Deep Think
  4. 72.9 Nemotron-Cascade 2 30B-A3B
  5. 71.9 GPT-5.5
  6. 61.9 DeepSeekMath-V2
  7. 54.3 Seed2.1 Pro
  8. 49.0 Gemini 3.1 Pro
  9. 38.4 Intern-S2-397B
  10. 35.7 GPT-5.2
  11. 30.0 Gemini 3 Pro
  12. 28.6 GPT-5 Pro
  13. 23.8 Claude Opus 4.5
  14. 23.3 Grok 4 Heavy
  15. 20.5 o3
  16. 20.0 GPT-5
  17. 18.6 Grok 4
  18. 18.6 Grok 4.1 Fast
  19. 17.6 Gemini 2.5 Pro
  20. 11.4 o4-mini
  21. 7.10 GPT-5.1
  22. 4.80 Claude Sonnet 4
  23. 4.30 DeepSeek-V3
  24. 3.80 DeepSeek-R1
  25. 2.90 Claude Opus 4
  1. 1 GPT-5.5 Pro xhigh 88.1%
  2. 2 Nemotron 3 Ultra 82.3%
  3. 3 Gemini 3 Deep Think 76.7%
  4. 4 Nemotron-Cascade 2 30B-A3B 72.9%
  5. 5 GPT-5.5 xhigh 71.9%
  6. 6 DeepSeekMath-V2 61.9%
  7. 7 Seed2.1 Pro 54.3%
  8. 8 Gemini 3.1 Pro 49%
  9. 9 Intern-S2-397B 38.4%
  10. 10 GPT-5.2 high 35.7%
  11. 11 Gemini 3 Pro 30%
  12. 12 GPT-5 Pro 28.6%
  13. 13 Claude Opus 4.5 23.8%
  14. 14 Grok 4 Heavy 23.3%
  15. 15 o3 20.5%
  16. 16 GPT-5 20%
  17. 17 Grok 4 18.6%
  18. 18 Grok 4.1 Fast reasoning 18.6%
  19. 19 Gemini 2.5 Pro Jun 2025 17.6%
  20. 20 o4-mini high 11.4%
  21. 21 GPT-5.1 7.10%
  22. 22 Claude Sonnet 4 4.80%
  23. 23 DeepSeek-V3 Mar 2025 4.30%
  24. 24 DeepSeek-R1 May 2025 3.80%
  25. 25 Claude Opus 4 2.90%

Sources: Official IMO-ProofBench leaderboard and labs’ model cards

#EntryScoreSettingDateSource
1 Aletheia 91.9% Feb 2026 Official leaderboard
2 GPT-5.5 Pro 88.1% xhigh Apr 2026 Official leaderboard
3 Gemini 3 Deep Think 76.7% Feb 2026 Official leaderboard
4 GPT-5.5 71.9% xhigh Apr 2026 Official leaderboard
5 Gemini Deep Think (IMO Gold) 65.7% Aug 2025 Official leaderboard
6 Gemini 3.1 Pro 49% Feb 2026 Official leaderboard
7 Gemini Deep Think (IMO lite) 37.6% Aug 2025 Official leaderboard
8 GPT-5.2 35.7% high Dec 2025 Official leaderboard
9 Gemini 3 Pro 30% Nov 2025 Official leaderboard
10 GPT-5 Pro 28.6% Oct 2025 Official leaderboard
11 Gemini 2.5 Pro with ( Huang & Yang, 2025 ) 24.8% Jul 2025 Official leaderboard
12 Claude Opus 4.5 23.8% Nov 2025 Official leaderboard
13 Grok 4 Heavy 23.3% Jul 2025 Official leaderboard
14 o3 20.5% Apr 2025 Official leaderboard
15 GPT-5 20% Aug 2025 Official leaderboard
16 Grok 4 18.6% Jul 2025 Official leaderboard
17 Grok 4.1 Fast 18.6% reasoning Nov 2025 Official leaderboard
18 Gemini 2.5 Pro Jun 2025 17.6% Jun 2025 Official leaderboard
19 o4-mini 11.4% high Apr 2025 Official leaderboard
20 Kimi-K2-Instruct 7.10% Aug 2025 Official leaderboard
21 GPT-5.1 7.10% Nov 2025 Official leaderboard
22 Qwen3-235B 5.20% Aug 2025 Official leaderboard
23 Claude Sonnet 4 4.80% May 2025 Official leaderboard
24 DeepSeek-V3 Mar 2025 4.30% Mar 2025 Official leaderboard
25 DeepSeek-R1 May 2025 3.80% May 2025 Official leaderboard
26 Claude Opus 4 2.90% May 2025 Official leaderboard
#ModelScoreSettingDateSource
1 GPT-5.5 Pro 88.1% xhigh Apr 2026 Official leaderboard
2 Nemotron 3 Ultra 82.3% Jun 2026 Lab model card
3 Gemini 3 Deep Think 76.7% Feb 2026 Official leaderboard
4 Nemotron-Cascade 2 30B-A3B 72.9% Mar 2026 Lab model card
5 GPT-5.5 71.9% xhigh Apr 2026 Official leaderboard
6 DeepSeekMath-V2 61.9% Nov 2025 Lab model card
7 Seed2.1 Pro 54.3% Jun 2026 Lab model card
8 Gemini 3.1 Pro 49% Feb 2026 Official leaderboard
9 Intern-S2-397B 38.4% Sep 2026 Lab model card
10 GPT-5.2 35.7% high Dec 2025 Official leaderboard
11 Gemini 3 Pro 30% Nov 2025 Official leaderboard
12 GPT-5 Pro 28.6% Oct 2025 Official leaderboard
13 Claude Opus 4.5 23.8% Nov 2025 Official leaderboard
14 Grok 4 Heavy 23.3% Jul 2025 Official leaderboard
15 o3 20.5% Apr 2025 Official leaderboard
16 GPT-5 20% Aug 2025 Official leaderboard
17 Grok 4 18.6% Jul 2025 Official leaderboard
18 Grok 4.1 Fast 18.6% reasoning Nov 2025 Official leaderboard
19 Gemini 2.5 Pro Jun 2025 17.6% Jun 2025 Official leaderboard
20 o4-mini 11.4% high Apr 2025 Official leaderboard
21 GPT-5.1 7.10% Nov 2025 Official leaderboard
22 Claude Sonnet 4 4.80% May 2025 Official leaderboard
23 DeepSeek-V3 Mar 2025 4.30% Mar 2025 Official leaderboard
24 DeepSeek-R1 May 2025 3.80% May 2025 Official leaderboard
25 Claude Opus 4 2.90% May 2025 Official leaderboard

Labs report scores for 5 models in their cards. Switch the scores above to rank them together.

5 models have held number one on the official leaderboard since its release in November 2025 5 models have held number one with claims included since its release in November 2025

Lab claims count from the day the card came out, independent runs from the model’s release. We draw one row each for the benchmark’s own leaderboard and what labs claim in their cards.

OpenAI Google DeepMind DeepSeek NVIDIA
Own leaderboard Passed 85% in Apr 2026, by GPT-5.5 Pro
GPT-5.5 Pro 88.1%
Lab claims Not tested since Sep 2026
Nemotron 3 Ultra 82.3%

Sources: Official IMO-ProofBench leaderboard and labs’ model cards

GPT-5.5 Pro on the official leaderboard passed 85% in April 2026

All the scores we have, one dot each, so a model tested at several settings appears more than once.

On the benchmark’s own leaderboard Reported by the lab in its model card Best so far
0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) GPT-5.5 Pro Gemini 3 Deep Think DeepSeekMath-V2 Gemini 3 Pro Grok 4 Heavy DeepSeek-V3 0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) GPT-5.5 Pro Gemini 3 Deep Think

Sources: Official IMO-ProofBench leaderboard and labs’ model cards

Labs headline it less often now, last in September 2026

Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.

DeepSeek
NVIDIA
Shanghai AI Lab
DateLaunch postScoreAs the post writes it
Sep 13, 2026 Intern-S2-397B 38.4 IMO-Proof (Olympiad-Level Mathematics Proof)
Mar 19, 2026 Nemotron-Cascade 2 30B-A3B 72.9 IMO ProofBench
Nov 27, 2025 DeepSeekMath-V2 61.9 ProofBench-Advanced, DeepSeekMath-V2 (Heavy), human evaluations (Basic 99.0)

4 system cards report it, from 3 labs

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.

DateDocumentScoreAs the document writes itSection
Jun 23, 2026 Seed2.1 Pro model card 54.3 IMOProof-Adv, Seed2.1 Pro column (Table 7). Deep Think 83.8 Capabilities
Jun 9, 2026 Nemotron 3 Ultra technical report 82.3 IMO-ProofBench Advanced, 173/210, generate-verify-refine test-time scaling, human graded (Table 11) Capabilities
Mar 19, 2026 Nemotron-Cascade 2 30B-A3B technical report 72.9 IMO Proof Bench Overall (60), generate-verify-refine test-time scaling, DeepSeek-V3.2-Speciale judge (Tables 1 and 5). Basic 92.5, Advanced 53.4 Capabilities
Nov 27, 2025 DeepSeekMath-V2 technical report 61.9 ProofBench-Advanced, DeepSeekMath-V2 (Heavy), expert evaluation (Figure 3). ProofBench-Basic 99.0 Capabilities

More in Math

All 113 in Math
  • 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%
  • IMO-AnswerBench In use 400 olympiad problems with short verifiable answers, robustified against memorization. Scored by answer match. GLM-5.2 91%
  • 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%
  • 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%
  • 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%
  • MathArena Apex Shortlist In use Models solve 47 hard final-answer problems that MathArena selected from 2025 competitions, where the best models scored about 50% at selection. Scored by accuracy over repeated attempts. GPT-5.5 98.4%