MATH (competition mathematics)
Integrity rank #50 of 65 · 85 models scored · top score 98.13 · GPT-5
Competition math; L5=hardest band
Strongest on discrimination, weakest on contamination resistance. Caveat: the public test set does not rule out training-data contamination.
Reported scores — protocols may differ. How we compare
| # | Model | Score | Evidence | Measured |
|---|---|---|---|---|
| 1 | GPT-5 | 98.13 | reproduced· optimizedT1 | 2025-08-07 |
| 2 | GPT-5 mini | 97.85 | reproduced· optimizedT1 | 2025-08-07 |
| 3 | o4-mini | 97.83 | reproduced· optimizedT1 | 2025-04-16 |
| 4 | o3 | 97.77 | reproduced· optimizedT1 | 2025-04-16 |
| 5 | Claude Sonnet 4.5 | 97.73 | reproduced· optimizedT1 | 2025-09-29 |
| 6 | Qwen3-Max | 97.13 | reproduced· optimizedT1 | 2025-09-24 |
| 7 | DeepSeek-R1 (May 2025) | 96.64 | reproduced· optimizedT1 | 2025-05-28 |
| 8 | o3-mini | 96.49 | reproduced· optimizedT1 | 2025-01-31 |
| 9 | Claude Haiku 4.5 | 96.36 | reproduced· optimizedT1 | 2025-10-15 |
| 10 | Gemini 2.5 Pro (May 2025) | 95.9 | reproduced· optimizedT1 | 2025-05-06 |
| 11 | Gemini 2.5 Pro (Mar 2025) | 95.56 | reproduced· optimizedT1 | 2025-03-31 |
| 12 | GPT-5 nano | 95.24 | reproduced· optimizedT1 | 2025-08-07 |
| 13 | o1 | 94.71 | reproduced· optimizedT1 | 2024-12-17 |
| 14 | DeepSeek-R1 | 93.05 | reproduced· optimizedT1 | 2025-01-20 |
| 15 | Claude 3.7 Sonnet | 91.16 | reproduced· optimizedT1 | 2025-02-24 |
| 16 | Grok-3 mini | 90.94 | reproduced· optimizedT1 | 2025-04-09 |
| 17 | o1-mini | 89.18 | reproduced· optimizedT1 | 2024-09-12 |
| 18 | Grok 3 | 88.75 | reproduced· optimizedT1 | 2025-04-09 |
| 19 | GPT-4.1 mini | 87.29 | reproduced· optimizedT1 | 2025-04-14 |
| 20 | DeepSeek-R1-Distill-Qwen-14B | 87.12 | unverifiedT2 | 2025-01-20 |
| 21 | Claude Opus 4 | 85.05 | reproduced· optimizedT1 | 2025-05-22 |
| 22 | Claude Sonnet 4 | 84.37 | reproduced· optimizedT1 | 2025-05-22 |
| 23 | Gemini 2.0 Pro | 83.46 | reproduced· optimizedT1 | 2025-02-05 |
| 24 | GPT-4.1 | 83.01 | reproduced· optimizedT1 | 2025-04-14 |
| 25 | Gemini 2.0 Flash (Feb 2025) | 82.17 | reproduced· optimizedT1 | 2025-02-05 |
| 26 | o1-preview | 81.65 | reproduced· optimizedT1 | 2024-09-12 |
| 27 | Mistral Medium 3 | 81.63 | reproduced· optimizedT1 | 2025-05-07 |
| 28 | GPT-4.5 | 78.63 | reproduced· optimizedT1 | 2025-02-27 |
| 29 | DeepSeek-V3 (Mar 2025) | 75.55 | reproduced· optimizedT1 | 2025-03-24 |
| 30 | Gemma 3 27B | 74.04 | reproduced· optimizedT1 | 2025-03-12 |
| 31 | Llama 4 Maverick | 73.02 | reproduced· optimizedT1 | 2025-04-05 |
| 32 | Gemini 1.5 Pro (Sept 2024) | 70.39 | reproduced· optimizedT1 | 2024-09-24 |
| 33 | GPT-4.1 nano | 70 | reproduced· optimizedT1 | 2025-04-14 |
| 34 | Qwen3-235B-A22B | 68.86 | reproduced· optimizedT1 | 2025-04-28 |
| 35 | Qwen2.5-Max | 67.18 | reproduced· optimizedT1 | 2025-01-25 |
| 36 | Phi-4 | 64.94 | reproduced· optimizedT1 | 2024-12-12 |
| 37 | DeepSeek-V3 | 64.85 | reproduced· optimizedT1 | 2024-12-26 |
| 38 | Grok-2 (Dec 2024) | 63.52 | reproduced· optimizedT1 | 2024-12-12 |
| 39 | Qwen2.5-72B | 63.17 | reproduced· optimizedT1 | 2024-09-19 |
| 40 | Llama 4 Scout | 62.27 | reproduced· optimizedT1 | 2025-04-05 |
| 41 | Gemini 1.5 Flash (Sep 2024) | 61.87 | reproduced· optimizedT1 | 2024-09-24 |
| 42 | Claude 3.5 Sonnet (October 2024) | 56.95 | reproduced· optimizedT1 | 2024-10-22 |
| 43 | Qwen2.5-32B | 56.07 | unverifiedT2 | 2024-09-17 |
| 44 | GPT-4o (Aug 2024) | 53.28 | reproduced· optimizedT1 | 2024-08-06 |
| 45 | GPT-4o mini | 52.63 | reproduced· optimizedT1 | 2024-07-18 |
| 46 | Claude 3.5 Sonnet | 51.68 | reproduced· optimizedT1 | 2024-06-20 |
| 47 | GPT-4o (May 2024) | 51.05 | reproduced· optimizedT1 | 2024-05-13 |
| 48 | Mistral Large 2 (Nov 2024) | 50.28 | reproduced· optimizedT1 | 2024-11-18 |
| 49 | Llama 3.1-405B | 49.77 | reproduced· optimizedT1 | 2024-07-23 |
| 50 | GPT-4o (Nov 2024) | 49.77 | reproduced· optimizedT1 | 2024-11-20 |
| 51 | Mistral Small 3.1 | 46.77 | reproduced· optimizedT1 | 2025-03-17 |
| 52 | GPT-4 Turbo (Apr 2024) | 46.73 | reproduced· optimizedT1 | 2024-04-09 |
| 53 | Claude 3.5 Haiku | 46.36 | reproduced· optimizedT1 | 2024-10-22 |
| 54 | Mistral Large 2 (Jul 2024) | 44.82 | reproduced· optimizedT1 | 2024-07-24 |
| 55 | Mistral Small 3 | 44.82 | unverifiedT2 | 2025-01-25 |
| 56 | Llama 3.3 70B | 41.6 | reproduced· optimizedT1 | 2024-12-06 |
| 57 | Gemini 1.5 Pro (May 2024) | 40.75 | reproduced· optimizedT1 | 2024-05-14 |
| 58 | GPT-4 Turbo (Nov 2023) | 40.02 | reproduced· optimizedT1 | 2024-01-25 |
| 59 | Llama 3.2 90B | 39.44 | reproduced· optimizedT1 | 2024-09-24 |
| 60 | Qwen2-72B | 39.07 | reproduced· optimizedT1 | 2024-06-07 |
| 61 | Claude 3 Opus | 37.48 | reproduced· optimizedT1 | 2024-02-29 |
| 62 | Llama 3.1-70B | 36.68 | reproduced· optimizedT1 | 2024-07-23 |
| 63 | Gemma 2 27B | 27.89 | reproduced· optimizedT1 | 2024-06-24 |
| 64 | Gemini 1.5 Flash (May 2024) | 25.12 | reproduced· optimizedT1 | 2024-05-23 |
| 65 | Mistral Large | 24.46 | reproduced· optimizedT1 | 2024-02-26 |
| 66 | Mixtral 8x22B | 24.24 | reproduced· optimizedT1 | 2024-04-17 |
| 67 | GPT-4 (Jun 2023) | 22.97 | reproduced· optimizedT1 | 2023-06-13 |
| 68 | Llama 3.1-8B | 22.88 | reproduced· optimizedT1 | 2024-07-23 |
| 69 | Llama 3-70B | 22.55 | reproduced· optimizedT1 | 2024-04-18 |
| 70 | Gemma 2 9B | 21.01 | reproduced· optimizedT1 | 2024-06-24 |
| 71 | Claude 3 Sonnet | 18.17 | reproduced· optimizedT1 | 2024-02-29 |
| 72 | phi-3-medium 14B | 17.56 | reproduced· optimizedT1 | 2024-04-23 |
| 73 | GPT-3.5 Turbo (Nov 2023) | 15.89 | reproduced· optimizedT1 | 2023-11-06 |
| 74 | Claude 3 Haiku | 14.88 | reproduced· optimizedT1 | 2024-03-07 |
| 75 | Ministral 3B | 14.44 | unverifiedT2 | 2024-10-16 |
| 76 | Claude 2 | 11.73 | reproduced· optimizedT1 | 2023-07-11 |
| 77 | GPT-3.5 Turbo (Jan 2024) | 11.63 | reproduced· optimizedT1 | 2024-01-25 |
| 78 | Gemini 1.0 Pro | 11.24 | reproduced· optimizedT1 | 2023-12-13 |
| 79 | Mistral NeMo | 10.83 | reproduced· optimizedT1 | 2024-07-18 |
| 80 | Mixtral 8x7B | 9.95 | reproduced· optimizedT1 | 2023-12-11 |
| 81 | DeepSeek LLM 67B | 6.39 | unverifiedT2 | 2023-11-29 |
| 82 | Llama 3-8B | 6.13 | reproduced· optimizedT1 | 2024-04-18 |
| 83 | Yi-34B | 5.15 | reproduced· optimizedT1 | 2023-11-02 |
| 84 | Mistral 7B v0.3 | 3.68 | unverifiedT2 | 2024-05-27 |
| 85 | Llama 2-70B | 3.29 | reproduced· optimizedT1 | 2023-07-18 |
Continue from this leaderboard into the tightest head-to-head reads.
Weighted composite. Each dimension opens to the source-backed sub-signals it is computed from.
Claims drawn from cited facts, not live model generation.
This benchmark scores 85 models, and the gap between its best and worst performers spans 94.84 points. A spread that wide indicates substantial separation between models, so differences in rank order on this benchmark are likely to reflect genuine capability differences rather than noise.
3 cited facts
This benchmark's Benchmark Integrity Index is 73, graded B, ranking 48 out of 62 benchmarks. The weakest component is contamination, which narrows how much to trust top scores. This score reflects the benchmark's health as a discriminator of frontier models under a disclosed harness, not any model's capability.
6 cited facts
This benchmark is saturated, with a top score of 98.13 and only 1.87 points of headroom remaining to the ceiling. Nine models cluster within a couple of points of that top mark. That means today's ranking among the top models is close to noise, so small differences there should not be read as real capability gaps.
5 cited facts
This benchmark is scored under a consistent harness, so reported results are directly comparable across evaluations that follow the same protocol, and they should be read as task performance under that disclosed harness rather than as evidence of deployed capability. Because the test set is public and has a documented history of heavy contamination and memorization, strong results warrant particular scrutiny for possible leakage rather than being taken at face value, unlike a held-out set where such concerns would be reduced. The harness itself is sensitive to configuration, with enhanced decoding and tool use raising scores markedly relative to a plain single-shot setup, so differences in harness details can shift outcomes even when the underlying task is unchanged.
4 cited facts
This benchmark measures competition mathematics, including a designated hardest difficulty band, so it targets advanced problem-solving rather than broad language competence. Its test set was assembled from competition archives with difficulty tags, making the items curated rather than newly authored for the benchmark. The central assumption is that matching the final answer reflects correct reasoning, treating answer accuracy as a proxy for a valid solution process. The sharpest caveat is that the benchmark can reward guessed answers, all items are public, and a code interpreter can trivialize many problems, so high scores may overstate genuine reasoning ability.
4 cited facts
MATH level 5 (MATH (competition mathematics)): MATH Level 5 — the hardest (level-5) competition-mathematics problems from the MATH dataset; scored as accuracy (%).
GPT-5 leads MATH level 5 at 98.13 — independently reproduced. The full leaderboard above lists every recorded measurement, not just the headline number.
85 models have recorded scores on MATH level 5, spanning a score spread of 94.84.
tensor.news grades MATH level 5 B for integrity (score 73/100), ranking #50 of 65 benchmarks we assess across discrimination, saturation, contamination resistance, harness comparability, and freshness.
Yes — top scores are clustering near the ceiling (saturation ratio 0.11), so MATH level 5 no longer separates leading models well. Treat small gaps at the top with caution.
Its test set is public, so contamination risk is on the table: a high score may partly reflect training-data overlap rather than capability.
Scores are reported under a consistent harness, so comparisons on MATH level 5 are reasonably apples-to-apples.
Every MATH level 5 measurement is source-backed and tagged reproduced, self-reported, or unverified; the underlying record cites arxiv.org/abs/2103.03874.