IMO-AnswerBench
The benchmark’s paper hasn’t scored a model released after GPT-5 (August 2025). Labs report their own numbers for 37 newer models, such as GLM-5.2 at 91%.
400 olympiad problems with short verifiable answers, robustified against memorization. Scored by answer match.
Part of IMO-Bench (EMNLP 2025) with IMO-ProofBench, IMO-GradingBench and IMO-LeanProofBench. On February 12th, 2026 a v2 file fixed ambiguous problems and wrong answers in place, so check which file a score used.
- Made by
- Google DeepMind
- Released
- November 2025
- Announced on X
- 188K views, 694 likes
- Kind
- Competition problems with answers
- Level
- Competition
- Scored
- Checked automatically
- Answers
- Short answer
- Tasks
- All public
- Measures
- Capability
- Subject
- Mathematics
- Language
- English
- Interaction
- Single turn
- Items from
- Exams and contests
- Refresh
- New versions replace it
- Run by
- Anyone
- Based on
- A version of another benchmark
- Strongest model tested
-
Qwen3.7-Max #49 on the capability index
- Last in a launch post
- October 2026
- Last model added
- October 2026
- Scores refreshed
- November 18th, 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/.
GLM-5.2 leads with 91%
Best score per model as labs report them in their own cards, drawn striped. Labs pick their own settings, so these compare less cleanly than independent runs.
Grok 4 leads with 73.1% across all independent runs
Best score per model across the benchmark’s paper, Epoch AI’s own runs and any board Epoch AI copied. Dotted bars are reruns by Epoch AI or another board, and each row names its source.
GLM-5.2 leads with 91%, 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. Striped bars are claims, and dotted bars are reruns by Epoch AI or another board.
- 91.0 GLM-5.2
- 90.0
Ring-2.5-1T
- 90.0
Qwen3.7-Max
- 90.0
Tencent Hy3
- 89.8
DeepSeek-V4-Pro
- 89.3
Seed2.0 Pro
- 88.6
Nemotron 3 Ultra
- 88.4
DeepSeek-V4-Flash
- 86.8
LongCat-Flash-Thinking-2601
- 86.1
Ring-2.6-1T
- 86.0 Kimi K2.6
- 86.0
Qwen3.7-Plus
- 85.4
Step 3.5 Flash
- 84.5
DeepSeek-V3.2-Speciale
- 84.3
Tencent Hy3 preview
- 83.9
Qwen3-Max-2026-01-23
- 83.8
Qwen 3.6 Plus
- 83.8 GLM-5.1
- 82.5 GLM-5
- 82.0 GLM-4.7
- 81.8 Kimi K2.5
- 81.8
LongCat-2.0
- 80.9
Qwen3.5 397B-A17B
- 80.8
Qwen3.6 27B
- 79.3
Nemotron-Cascade 2 30B-A3B
- 78.9
Qwen 3.6 35B-A3B
- 78.6 Kimi K2 Thinking
- 78.6
K-EXAONE 2.0
- 78.3
DeepSeek-V3.2
- 77.3
Intern-S1-Pro
- 76.3
Seed1.8
- 76.3
K-EXAONE
- 75.2
Spark-X2.6-4B
- 73.4
Tencent HY 2.0 Think
- 65.8
Ling-2.6-1T
- 65.5
Qwen3.5-Omni-Plus
- 62.1
STEP3-VL-10B
- 1 GLM-5.2 91%
- 2
Ring-2.5-1T 90%
- 3
Qwen3.7-Max 90%
- 4
Tencent Hy3 90%
- 5
DeepSeek-V4-Pro 89.8%
- 6
Seed2.0 Pro 89.3%
- 7
Nemotron 3 Ultra 88.6%
- 8
DeepSeek-V4-Flash 88.4%
- 9
LongCat-Flash-Thinking-2601 86.8%
- 10
Ring-2.6-1T 86.1%
- 11 Kimi K2.6 86%
- 12
Qwen3.7-Plus 86%
- 13
Step 3.5 Flash 85.4%
- 14
DeepSeek-V3.2-Speciale 84.5%
- 15
Tencent Hy3 preview 84.3%
- 16
Qwen3-Max-2026-01-23 83.9%
- 17
Qwen 3.6 Plus 83.8%
- 18 GLM-5.1 83.8%
- 19 GLM-5 82.5%
- 20 GLM-4.7 82%
- 21 Kimi K2.5 81.8%
- 22
LongCat-2.0 81.8%
- 23
Qwen3.5 397B-A17B 80.9%
- 24
Qwen3.6 27B 80.8%
- 25
Nemotron-Cascade 2 30B-A3B 79.3%
- 26
Qwen 3.6 35B-A3B 78.9%
- 27 Kimi K2 Thinking 78.6%
- 28
K-EXAONE 2.0 78.6%
- 29
DeepSeek-V3.2 78.3%
- 30
Intern-S1-Pro 77.3%
- 31
Seed1.8 76.3%
- 32
K-EXAONE 76.3%
- 33
Spark-X2.6-4B 75.2%
- 34
Tencent HY 2.0 Think 73.4%
- 35
Ling-2.6-1T 65.8%
- 36
Qwen3.5-Omni-Plus 65.5%
- 37
STEP3-VL-10B 62.1%
Source: Labs’ model cards
- 1 Grok 4 73.1%
- 2
Gemini 2.5 Deep Think 71.8%
- 3 o4-mini high reasoning 67.9%
- 4 GPT-5 65.6%
- 5 o3 61.1%
- 6
Qwen3-235B-A22B-Instruct Jul 2025 53.8%
- 7 Kimi K2 Jul 2025 45.8%
- 8 Claude Sonnet 4 23%
- 9 Claude Opus 4 22.3%
Source: The IMO-AnswerBench paper
- 91.0 GLM-5.2
- 90.0
Ring-2.5-1T
- 90.0
Qwen3.7-Max
- 90.0
Tencent Hy3
- 89.8
DeepSeek-V4-Pro
- 89.3
Seed2.0 Pro
- 88.6
Nemotron 3 Ultra
- 88.4
DeepSeek-V4-Flash
- 86.8
LongCat-Flash-Thinking-2601
- 86.1
Ring-2.6-1T
- 86.0 Kimi K2.6
- 86.0
Qwen3.7-Plus
- 85.4
Step 3.5 Flash
- 84.5
DeepSeek-V3.2-Speciale
- 84.3
Tencent Hy3 preview
- 83.9
Qwen3-Max-2026-01-23
- 83.8
Qwen 3.6 Plus
- 83.8 GLM-5.1
- 82.5 GLM-5
- 82.0 GLM-4.7
- 81.8 Kimi K2.5
- 81.8
LongCat-2.0
- 80.9
Qwen3.5 397B-A17B
- 80.8
Qwen3.6 27B
- 79.3
Nemotron-Cascade 2 30B-A3B
- 78.9
Qwen 3.6 35B-A3B
- 78.6 Kimi K2 Thinking
- 78.6
K-EXAONE 2.0
- 78.3
DeepSeek-V3.2
- 77.3
Intern-S1-Pro
- 76.3
Seed1.8
- 76.3
K-EXAONE
- 75.2
Spark-X2.6-4B
- 73.4
Tencent HY 2.0 Think
- 73.1 Grok 4
- 71.8
Gemini 2.5 Deep Think
- 67.9 o4-mini
- 65.8
Ling-2.6-1T
- 65.6 GPT-5
- 65.5
Qwen3.5-Omni-Plus
- 62.1
STEP3-VL-10B
- 61.1 o3
- 53.8
Qwen3-235B-A22B-Instruct
- 45.8 Kimi K2
- 23.0 Claude Sonnet 4
- 22.3 Claude Opus 4
- 1 GLM-5.2 91%
- 2
Ring-2.5-1T 90%
- 3
Qwen3.7-Max 90%
- 4
Tencent Hy3 90%
- 5
DeepSeek-V4-Pro 89.8%
- 6
Seed2.0 Pro 89.3%
- 7
Nemotron 3 Ultra 88.6%
- 8
DeepSeek-V4-Flash 88.4%
- 9
LongCat-Flash-Thinking-2601 86.8%
- 10
Ring-2.6-1T 86.1%
- 11 Kimi K2.6 86%
- 12
Qwen3.7-Plus 86%
- 13
Step 3.5 Flash 85.4%
- 14
DeepSeek-V3.2-Speciale 84.5%
- 15
Tencent Hy3 preview 84.3%
- 16
Qwen3-Max-2026-01-23 83.9%
- 17
Qwen 3.6 Plus 83.8%
- 18 GLM-5.1 83.8%
- 19 GLM-5 82.5%
- 20 GLM-4.7 82%
- 21 Kimi K2.5 81.8%
- 22
LongCat-2.0 81.8%
- 23
Qwen3.5 397B-A17B 80.9%
- 24
Qwen3.6 27B 80.8%
- 25
Nemotron-Cascade 2 30B-A3B 79.3%
- 26
Qwen 3.6 35B-A3B 78.9%
- 27 Kimi K2 Thinking 78.6%
- 28
K-EXAONE 2.0 78.6%
- 29
DeepSeek-V3.2 78.3%
- 30
Intern-S1-Pro 77.3%
- 31
Seed1.8 76.3%
- 32
K-EXAONE 76.3%
- 33
Spark-X2.6-4B 75.2%
- 34
Tencent HY 2.0 Think 73.4%
- 35 Grok 4 73.1%
- 36
Gemini 2.5 Deep Think 71.8%
- 37 o4-mini high reasoning 67.9%
- 38
Ling-2.6-1T 65.8%
- 39 GPT-5 65.6%
- 40
Qwen3.5-Omni-Plus 65.5%
- 41
STEP3-VL-10B 62.1%
- 42 o3 61.1%
- 43
Qwen3-235B-A22B-Instruct Jul 2025 53.8%
- 44 Kimi K2 Jul 2025 45.8%
- 45 Claude Sonnet 4 23%
- 46 Claude Opus 4 22.3%
Sources: Labs’ model cards and the IMO-AnswerBench paper
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | GLM-5.2 | 91% | Jun 2026 | Lab model card | |
| 2 | | 90% | Feb 2026 | Lab model card | |
| 3 | | 90% | May 2026 | Lab model card | |
| 4 | | 90% | Jul 2026 | Lab model card | |
| 5 | | 89.8% | Apr 2026 | Lab model card | |
| 6 | | 89.3% | Feb 2026 | Lab model card | |
| 7 | | 88.6% | Jun 2026 | Lab model card | |
| 8 | | 88.4% | Apr 2026 | Lab model card | |
| 9 | | 86.8% | Jan 2026 | Lab model card | |
| 10 | | 86.1% | Jun 2026 | Lab model card | |
| 11 | Kimi K2.6 | 86% | Apr 2026 | Lab model card | |
| 12 | | 86% | Jun 2026 | Lab model card | |
| 13 | | 85.4% | Feb 2026 | Lab model card | |
| 14 | | 84.5% | Dec 2025 | Lab model card | |
| 15 | | 84.3% | Apr 2026 | Lab model card | |
| 16 | | 83.9% | Jan 2026 | Lab model card | |
| 17 | | 83.8% | Apr 2026 | Lab model card | |
| 18 | GLM-5.1 | 83.8% | Apr 2026 | Lab model card | |
| 19 | GLM-5 | 82.5% | Feb 2026 | Lab model card | |
| 20 | GLM-4.7 | 82% | Dec 2025 | Lab model card | |
| 21 | Kimi K2.5 | 81.8% | Jan 2026 | Lab model card | |
| 22 | | 81.8% | Jul 2026 | Lab model card | |
| 23 | | 80.9% | Feb 2026 | Lab model card | |
| 24 | | 80.8% | Apr 2026 | Lab model card | |
| 25 | | 79.3% | Mar 2026 | Lab model card | |
| 26 | | 78.9% | Apr 2026 | Lab model card | |
| 27 | Kimi K2 Thinking | 78.6% | Nov 2025 | Lab model card | |
| 28 | | 78.6% | Jul 2026 | Lab model card | |
| 29 | | 78.3% | Dec 2025 | Lab model card | |
| 30 | | 77.3% | Feb 2026 | Lab model card | |
| 31 | | 76.3% | Dec 2025 | Lab model card | |
| 32 | | 76.3% | Dec 2025 | Lab model card | |
| 33 | | 75.2% | Oct 2026 | Lab model card | |
| 34 | | 73.4% | Dec 2025 | Lab model card | |
| 35 | | 65.8% | Jun 2026 | Lab model card | |
| 36 | | 65.5% | Mar 2026 | Lab model card | |
| 37 | | 62.1% | Jan 2026 | Lab model card |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Grok 4 | 73.1% | Jul 2025 | Official leaderboard | |
| 2 | 71.8% | Aug 2025 | Official leaderboard | ||
| 3 | o4-mini | 67.9% | high reasoning | Apr 2025 | Official leaderboard |
| 4 | GPT-5 | 65.6% | Aug 2025 | Official leaderboard | |
| 5 | o3 | 61.1% | Apr 2025 | Official leaderboard | |
| 6 | 53.8% | Jul 2025 | Official leaderboard | ||
| 7 | Kimi K2 Jul 2025 | 45.8% | Jul 2025 | Official leaderboard | |
| 8 | Claude Sonnet 4 | 23% | May 2025 | Official leaderboard | |
| 9 | Claude Opus 4 | 22.3% | May 2025 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | GLM-5.2 | 91% | Jun 2026 | Lab model card | |
| 2 | 90% | Feb 2026 | Lab model card | ||
| 3 | 90% | May 2026 | Lab model card | ||
| 4 | 90% | Jul 2026 | Lab model card | ||
| 5 | 89.8% | Apr 2026 | Lab model card | ||
| 6 | 89.3% | Feb 2026 | Lab model card | ||
| 7 | 88.6% | Jun 2026 | Lab model card | ||
| 8 | 88.4% | Apr 2026 | Lab model card | ||
| 9 | 86.8% | Jan 2026 | Lab model card | ||
| 10 | 86.1% | Jun 2026 | Lab model card | ||
| 11 | Kimi K2.6 | 86% | Apr 2026 | Lab model card | |
| 12 | 86% | Jun 2026 | Lab model card | ||
| 13 | 85.4% | Feb 2026 | Lab model card | ||
| 14 | 84.5% | Dec 2025 | Lab model card | ||
| 15 | 84.3% | Apr 2026 | Lab model card | ||
| 16 | 83.9% | Jan 2026 | Lab model card | ||
| 17 | 83.8% | Apr 2026 | Lab model card | ||
| 18 | GLM-5.1 | 83.8% | Apr 2026 | Lab model card | |
| 19 | GLM-5 | 82.5% | Feb 2026 | Lab model card | |
| 20 | GLM-4.7 | 82% | Dec 2025 | Lab model card | |
| 21 | Kimi K2.5 | 81.8% | Jan 2026 | Lab model card | |
| 22 | 81.8% | Jul 2026 | Lab model card | ||
| 23 | 80.9% | Feb 2026 | Lab model card | ||
| 24 | 80.8% | Apr 2026 | Lab model card | ||
| 25 | 79.3% | Mar 2026 | Lab model card | ||
| 26 | 78.9% | Apr 2026 | Lab model card | ||
| 27 | Kimi K2 Thinking | 78.6% | Nov 2025 | Lab model card | |
| 28 | 78.6% | Jul 2026 | Lab model card | ||
| 29 | 78.3% | Dec 2025 | Lab model card | ||
| 30 | 77.3% | Feb 2026 | Lab model card | ||
| 31 | 76.3% | Dec 2025 | Lab model card | ||
| 32 | 76.3% | Dec 2025 | Lab model card | ||
| 33 | 75.2% | Oct 2026 | Lab model card | ||
| 34 | 73.4% | Dec 2025 | Lab model card | ||
| 35 | Grok 4 | 73.1% | Jul 2025 | Official leaderboard | |
| 36 | 71.8% | Aug 2025 | Official leaderboard | ||
| 37 | o4-mini | 67.9% | high reasoning | Apr 2025 | Official leaderboard |
| 38 | 65.8% | Jun 2026 | Lab model card | ||
| 39 | GPT-5 | 65.6% | Aug 2025 | Official leaderboard | |
| 40 | 65.5% | Mar 2026 | Lab model card | ||
| 41 | 62.1% | Jan 2026 | Lab model card | ||
| 42 | o3 | 61.1% | Apr 2025 | Official leaderboard | |
| 43 | 53.8% | Jul 2025 | Official leaderboard | ||
| 44 | Kimi K2 Jul 2025 | 45.8% | Jul 2025 | Official leaderboard | |
| 45 | Claude Sonnet 4 | 23% | May 2025 | Official leaderboard | |
| 46 | Claude Opus 4 | 22.3% | May 2025 | Official leaderboard |
The benchmark's paper has 13 entries, 9 of them models we track. Switch the scores above to rank them together.
Rankings by part
We read 4 parts from the benchmark’s paper, each ranking models on one part of the test, such as Algebra, Combinatorics and Geometry. Gemini 2.5 Deep Think leads 2 of them, and Grok 4 leads the other 2. Arrows show how far a model moved from its overall rank among the models we track.
Algebra
9 models we track have a score here.
- 1
Gemini 2.5 Deep Think ↑1 78%
- 2 Grok 4 ↓1 75.5%
- 3 o4-mini 71.3%
- 4 GPT-5 69.9%
- 5 o3 62.8%
- 6
Qwen3-235B-A22B-Instruct Jul 2025 57.6%
- 7 Kimi K2 Jul 2025 45.6%
- 8 Claude Sonnet 4 20.6%
- 9 Claude Opus 4 19.4%
Combinatorics
9 models we track have a score here.
- 1 Grok 4 55.9%
- 2
Gemini 2.5 Deep Think 49%
- 3 o4-mini 46.6%
- 4 GPT-5 46.4%
- 5 o3 43%
- 6
Qwen3-235B-A22B-Instruct Jul 2025 37.5%
- 7 Kimi K2 Jul 2025 31.1%
- 8 Claude Opus 4 ↑1 20%
- 9 Claude Sonnet 4 ↓1 17.8%
Geometry
9 models we track have a score here.
- 1
Gemini 2.5 Deep Think ↑1 83%
- 2 Grok 4 ↓1 80.1%
- 3 o4-mini 78.4%
- 4 GPT-5 74.8%
- 5 o3 70.6%
- 6
Qwen3-235B-A22B-Instruct Jul 2025 57.6%
- 7 Kimi K2 Jul 2025 49.3%
- 8 Claude Sonnet 4 26%
- 9 Claude Opus 4 23.3%
Number theory
9 models we track have a score here.
- 1 Grok 4 80.9%
- 2
Gemini 2.5 Deep Think 77%
- 3 o4-mini 75.3%
- 4 GPT-5 71.2%
- 5 o3 68%
- 6
Qwen3-235B-A22B-Instruct Jul 2025 62.3%
- 7 Kimi K2 Jul 2025 56.9%
- 8 Claude Sonnet 4 27.6%
- 9 Claude Opus 4 26.6%
5 models have held number one in labs’ own numbers since its release in November 2025 Number one over time 6 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 paper and what labs claim in their cards.
Meituan’s own number for LongCat-Flash-Thinking-2601 passed 85% in January 2026
That was 2 months after its release. By the same age, 11% of the benchmarks from 2025 had passed 85%. All the scores we have, one dot each, so a model tested at several settings appears more than once.
Sources: Official IMO-AnswerBench leaderboard and labs’ model cards
7 labs put it in a launch post in the last six months
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.
| Date | Launch post | Score | As the post writes it |
|---|---|---|---|
| Oct 22, 2026 | 75.2 | IMO-AnswerBench | |
| Jul 31, 2026 | 78.6 | IMO Answer (IMO-AnswerBench) | |
| Jul 5, 2026 | 81.8 | IMO-AnswerBench | |
| Jul 2, 2026 | 90 | IMOAnswerBench | |
| Jun 16, 2026 | GLM-5.2 | 91 | IMOAnswerBench |
| Jun 4, 2026 | 88.6 | IMOAnswerBench (no tools); with tools 92.3 | |
| Jun 1, 2026 | 86 | IMOAnswerBench | |
| May 20, 2026 | 90 | IMOAnswerBench | |
| Apr 24, 2026 | 88.4 | IMOAnswerBench (Pass@1), V4-Flash Max | |
| Apr 24, 2026 | 89.8 | IMOAnswerBench (Pass@1) | |
| Apr 23, 2026 | 84.3 | IMO Answer Bench | |
| Apr 22, 2026 | 80.8 | IMOAnswerBench | |
| Apr 20, 2026 | Kimi K2.6 | 86 | IMO-AnswerBench |
| Apr 15, 2026 | 78.9 | IMOAnswerBench | |
| Apr 7, 2026 | GLM-5.1 | 83.8 | IMOAnswerBench |
| Apr 2, 2026 | 83.8 | IMOAnswerBench | |
| Mar 30, 2026 | 65.5 | IMOAnswerBench | |
| Mar 19, 2026 | 79.3 | IMO AnswerBench | |
| Feb 16, 2026 | 80.9 | IMOAnswerBench | |
| Feb 14, 2026 | 89.3 | IMOAnswerBench (no tool) | |
| Feb 11, 2026 | GLM-5 | 82.5 | IMOAnswerBench |
| Feb 10, 2026 | 90 | Best of IMOAnswerBench (Mean@8) 86.69 / Heavy Thinking 90.00 | |
| Feb 2, 2026 | 77.3 | IMO-Answer-Bench | |
| Feb 2, 2026 | 85.4 | IMOAnswerBench | |
| Jan 27, 2026 | Kimi K2.5 | 81.8 | IMO-AnswerBench |
| Jan 23, 2026 | 83.9 | IMOAnswerBench | |
| Jan 14, 2026 | 86.8 | Best of IMO-AnswerBench (Avg@4) 78.6 / Heavy Thinking 86.8 | |
| Dec 31, 2025 | 76.3 | IMO-AnswerBench | |
| Dec 22, 2025 | GLM-4.7 | 82 | IMOAnswerBench |
| Dec 18, 2025 | 76.3 | IMO-AnswerBench (w/ code tools) | |
| Dec 5, 2025 | 73.4 | IMO-AnswerBench | |
| Dec 1, 2025 | 78.3 | IMOAnswerBench (Pass@1) | |
| Nov 6, 2025 | Kimi K2 Thinking | 78.6 | IMO-AnswerBench, no tools |
20 system cards report it, from 11 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.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| Sep 10, 2026 | | IMO-AnswerBench (Pass@1) vs reasoning effort, chart only (Figure 12) | Capabilities | |
| Aug 5, 2026 | | 78.6 | IMO-AnswerBench (Table 6, Reasoning mode) | Capabilities |
| Jun 13, 2026 | | 86.1 | IMO-AnswerBench (Avg@8), xhigh (Table 6). High gives 66.44 | Capabilities |
| Jun 13, 2026 | | 65.8 | IMO-AnswerBench, Mean@8, CoT (Table 5) | Capabilities |
| Jun 9, 2026 | | 88.6 | IMOAnswerBench (no tools), 92.3 with tools (Table 10) | Capabilities |
| Apr 26, 2026 | | 89.8 | IMOAnswerBench (Pass@1), DS-V4-Pro Max (Table 6) | Capabilities |
| Apr 26, 2026 | | 88.4 | IMOAnswerBench (Pass@1), DeepSeek-V4-Flash Max (Table 7) | Capabilities |
| Apr 17, 2026 | | 65.5 | IMOAnswerBench (Table 4) | Capabilities |
| Mar 26, 2026 | | 77.3 | IMO-Answer-Bench, thinking (Table 2) | Capabilities |
| Mar 21, 2026 | | 76.3 | IMO-AnswerBench, w/ code tools (Table 1) | Capabilities |
| Mar 19, 2026 | | 79.3 | IMO AnswerBench (Table 1) | Capabilities |
| Feb 17, 2026 | GLM-5 technical report | 82.5 | IMO-AnswerBench (Table 7) | Capabilities |
| Feb 14, 2026 | | 89.3 | IMOAnswerBench (no tool) (Table 3) | Capabilities |
| Feb 11, 2026 | | 85.4 | IMO-AnswerBench (Table 5). PaCoRe 88.8, with Python 86.7 (Table 12) | Capabilities |
| Jan 30, 2026 | Kimi K2.5 technical report | 81.8 | IMO-AnswerBench (Table 4) | Capabilities |
| Jan 23, 2026 | | 86.8 | IMO-AnswerBench (Avg@4), w/ tools, heavy mode (Table 2). Default mode 78.6 | Capabilities |
| Jan 14, 2026 | | 62.1 | IMO-AnswerBench, SeRe (Tables 2 and 3). PaCoRe 76.66 | Capabilities |
| Jan 5, 2026 | | 76.3 | IMO-AnswerBench (Table 3, Reasoning mode). Non-reasoning 40.0 (Table 4) | Capabilities |
| Dec 1, 2025 | | 78.3 | IMOAnswerBench (Pass@1), DeepSeek-V3.2 Thinking (Table 2) | Capabilities |
| Dec 1, 2025 | | 84.5 | IMOAnswerBench (Pass@1), DeepSeek-V3.2-Speciale (Table 3) | Capabilities |
One logged change that moves scores
Changes by the maintainers, so a score from before a change may not compare with one from after it. Newest first, each with the maintainers' own source. The list may be incomplete. What we log.
- Questions corrected February 12th, 2026
The maintainers replaced the original file with answerbench_v2.csv to correct ambiguous questions and incorrect answers. Scores using the deprecated file can differ. Source
More 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%
- 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%
- Riemann-bench In use 25 private research-mathematics problems from professors and IMO medalists. Agents with tools, 100 runs per problem. GPT-5.6 Sol 83.2%
Data audit
All 9 checks pass. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 3 pass 2 notes
- Pass Independent runs Scores from the benchmark’s paper (13 entries), launch posts (33 scores) and system cards (19 scores).
- Pass Models scored 46 models we track, plus 4 other entries on its leaderboard.
- Pass Newest score The newest is from October 2026, for Spark-X2.6-4B.
- Note Frontier models None of the top 10 has a score, which is fine for a retired test. The strongest tested is Qwen3.7-Max (#49).
- Note Headroom Saturated. The best is 91%, past 85% since January 2026.
Leaderboard import 3 pass
- Pass Importer We read the board on arxiv.org: 9 rows for models we track and 4 other entries at the last read.
- Pass Last read November 18th, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.