# Mathematical Reasoning on AIME 2025 (Acc, Tpi, Tok, Avg. Acc)

Source: [Mathematical Reasoning on AIME 2025 (Acc, Tpi, Tok, Avg. Acc)](<https://www.sota2.com/research/sota/mathematical-reasoning-on-aime-2025-acc-tpi-tok-avg-acc>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-02-26T05:31:14.508Z

Evaluation of model performance on the American Invitational Mathematics Examination (AIME) 2025 dataset, reporting Accuracy, Tokens per Inference (Tpi), Tokens (Tok), and Average Accuracy.

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

| Scope | Recorded value |
| --- | --- |
| Task | Mathematical Reasoning |
| Dataset (including any stated split) | AIME 2025 |
| Metric | Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 16 of 16 |
| Coverage | All results currently returned for this benchmark |

Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol.

## Metrics

- [Accuracy](<https://www.sota2.com/research/sota/mathematical-reasoning-on-aime-2025-acc-tpi-tok-avg-acc.md?metric=Accuracy>): Higher is better; unit: PERCENT.

- [Tpi](<https://www.sota2.com/research/sota/mathematical-reasoning-on-aime-2025-acc-tpi-tok-avg-acc.md?metric=Tpi>): Lower is better; unit: SCALAR.

- [Total Tokens](<https://www.sota2.com/research/sota/mathematical-reasoning-on-aime-2025-acc-tpi-tok-avg-acc.md?metric=Total%20Tokens>): Lower is better; unit: SCALAR.

- [Average Accuracy](<https://www.sota2.com/research/sota/mathematical-reasoning-on-aime-2025-acc-tpi-tok-avg-acc.md?metric=Average%20Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| Qwen3-32B-SCALE | 71.25 | Not recorded | 692e72a3d6b0dd334293237a |
| QwQ-SCALE | 70.21 | Not recorded | 692e72a3d6b0dd334293237a |
| Qwen3-32B-InftyThink | 70 | Not recorded | 692e72a3d6b0dd334293237a |
| QwQ-InftyThink | 65.53 | Not recorded | 692e72a3d6b0dd334293237a |
| QwQ-CoT | 63.33 | Not recorded | 692e72a3d6b0dd334293237a |
| QwQ-Majority-Voting | 63.33 | Not recorded | 692e72a3d6b0dd334293237a |
| Distill-Llama-70B-SCALE | 62.08 | Not recorded | 692e72a3d6b0dd334293237a |
| Distill-Llama-70B-CoT | 61.37 | Not recorded | 692e72a3d6b0dd334293237a |
| Distill-Qwen-32B-Majority-Voting | 60 | Not recorded | 692e72a3d6b0dd334293237a |
| Distill-Qwen-32B-SCALE | 59.17 | Not recorded | 692e72a3d6b0dd334293237a |
| Distill-Qwen-32B-CoT | 58.75 | Not recorded | 692e72a3d6b0dd334293237a |
| Qwen3-32B-CoT | 57.5 | Not recorded | 692e72a3d6b0dd334293237a |
| Distill-Llama-70B-Majority-Voting | 56.67 | Not recorded | 692e72a3d6b0dd334293237a |
| Distill-Llama-70B-InftyThink | 54.58 | Not recorded | 692e72a3d6b0dd334293237a |
| Qwen3-32B-Majority-Voting | 53.33 | Not recorded | 692e72a3d6b0dd334293237a |
| Distill-Qwen-32B-InftyThink | 48.75 | Not recorded | 692e72a3d6b0dd334293237a |

### Qwen3-32B-SCALE

Accuracy: 71.25. Result ID: 6996d4b55698665ec4210e67.

[Reporting paper](<https://www.sota2.com/research/paper/692e72a3d6b0dd334293237a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/692e72a3d6b0dd334293237a.md>)

[Implementation](<https://github.com/XiaoYang66/DualThinking>)

Recorded evaluation configuration: {"Backbone":"Qwen3-32B","Strategy":"SCALE"}. This record does not assert independent reproduction.

### QwQ-SCALE

Accuracy: 70.21. Result ID: 6996d4b55698665ec4210e6b.

[Reporting paper](<https://www.sota2.com/research/paper/692e72a3d6b0dd334293237a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/692e72a3d6b0dd334293237a.md>)

[Implementation](<https://github.com/XiaoYang66/DualThinking>)

Recorded evaluation configuration: {"Backbone":"QwQ","Strategy":"SCALE"}. This record does not assert independent reproduction.

### Qwen3-32B-InftyThink

Accuracy: 70. Result ID: 6996d4b55698665ec4210e65.

[Reporting paper](<https://www.sota2.com/research/paper/692e72a3d6b0dd334293237a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7ba11e209131cea89728.md>)

Recorded evaluation configuration: {"Backbone":"Qwen3-32B","Strategy":"InftyThink"}. This record does not assert independent reproduction.

### QwQ-InftyThink

Accuracy: 65.53. Result ID: 6996d4b55698665ec4210e69.

[Reporting paper](<https://www.sota2.com/research/paper/692e72a3d6b0dd334293237a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7ba11e209131cea89728.md>)

Recorded evaluation configuration: {"Backbone":"QwQ","Strategy":"InftyThink"}. This record does not assert independent reproduction.

### QwQ-CoT

Accuracy: 63.33. Result ID: 6996d4b55698665ec4210e68.

[Reporting paper](<https://www.sota2.com/research/paper/692e72a3d6b0dd334293237a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6e571e209131cea3f27b.md>)

Recorded evaluation configuration: {"Backbone":"QwQ","Strategy":"Chain-of-Thought"}. This record does not assert independent reproduction.

### QwQ-Majority-Voting

Accuracy: 63.33. Result ID: 6996d4b55698665ec4210e6a.

[Reporting paper](<https://www.sota2.com/research/paper/692e72a3d6b0dd334293237a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6eb61e209131cea41b33.md>)

Recorded evaluation configuration: {"Backbone":"QwQ","Strategy":"Majority-Voting"}. This record does not assert independent reproduction.

### Distill-Llama-70B-SCALE

Accuracy: 62.08. Result ID: 6996d4b55698665ec4210e6f.

[Reporting paper](<https://www.sota2.com/research/paper/692e72a3d6b0dd334293237a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/692e72a3d6b0dd334293237a.md>)

[Implementation](<https://github.com/XiaoYang66/DualThinking>)

Recorded evaluation configuration: {"Backbone":"Distill-Llama-70B","Strategy":"SCALE"}. This record does not assert independent reproduction.

### Distill-Llama-70B-CoT

Accuracy: 61.37. Result ID: 6996d4b55698665ec4210e6c.

[Reporting paper](<https://www.sota2.com/research/paper/692e72a3d6b0dd334293237a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6e571e209131cea3f27b.md>)

Recorded evaluation configuration: {"Backbone":"Distill-Llama-70B","Strategy":"Chain-of-Thought"}. This record does not assert independent reproduction.

### Distill-Qwen-32B-Majority-Voting

Accuracy: 60. Result ID: 6996d4b55698665ec4210e72.

[Reporting paper](<https://www.sota2.com/research/paper/692e72a3d6b0dd334293237a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6eb61e209131cea41b33.md>)

Recorded evaluation configuration: {"Backbone":"Distill-Qwen-32B","Strategy":"Majority-Voting"}. This record does not assert independent reproduction.

### Distill-Qwen-32B-SCALE

Accuracy: 59.17. Result ID: 6996d4b55698665ec4210e73.

[Reporting paper](<https://www.sota2.com/research/paper/692e72a3d6b0dd334293237a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/692e72a3d6b0dd334293237a.md>)

[Implementation](<https://github.com/XiaoYang66/DualThinking>)

Recorded evaluation configuration: {"Backbone":"Distill-Qwen-32B","Strategy":"SCALE"}. This record does not assert independent reproduction.

### Distill-Qwen-32B-CoT

Accuracy: 58.75. Result ID: 6996d4b55698665ec4210e70.

[Reporting paper](<https://www.sota2.com/research/paper/692e72a3d6b0dd334293237a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6e571e209131cea3f27b.md>)

Recorded evaluation configuration: {"Backbone":"Distill-Qwen-32B","Strategy":"Chain-of-Thought"}. This record does not assert independent reproduction.

### Qwen3-32B-CoT

Accuracy: 57.5. Result ID: 6996d4b55698665ec4210e64.

[Reporting paper](<https://www.sota2.com/research/paper/692e72a3d6b0dd334293237a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6e571e209131cea3f27b.md>)

Recorded evaluation configuration: {"Backbone":"Qwen3-32B","Strategy":"Chain-of-Thought"}. This record does not assert independent reproduction.

### Distill-Llama-70B-Majority-Voting

Accuracy: 56.67. Result ID: 6996d4b55698665ec4210e6e.

[Reporting paper](<https://www.sota2.com/research/paper/692e72a3d6b0dd334293237a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6eb61e209131cea41b33.md>)

Recorded evaluation configuration: {"Backbone":"Distill-Llama-70B","Strategy":"Majority-Voting"}. This record does not assert independent reproduction.

### Distill-Llama-70B-InftyThink

Accuracy: 54.58. Result ID: 6996d4b55698665ec4210e6d.

[Reporting paper](<https://www.sota2.com/research/paper/692e72a3d6b0dd334293237a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7ba11e209131cea89728.md>)

Recorded evaluation configuration: {"Backbone":"Distill-Llama-70B","Strategy":"InftyThink"}. This record does not assert independent reproduction.

### Qwen3-32B-Majority-Voting

Accuracy: 53.33. Result ID: 6996d4b55698665ec4210e66.

[Reporting paper](<https://www.sota2.com/research/paper/692e72a3d6b0dd334293237a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6eb61e209131cea41b33.md>)

Recorded evaluation configuration: {"Backbone":"Qwen3-32B","Strategy":"Majority-Voting"}. This record does not assert independent reproduction.

### Distill-Qwen-32B-InftyThink

Accuracy: 48.75. Result ID: 6996d4b55698665ec4210e71.

[Reporting paper](<https://www.sota2.com/research/paper/692e72a3d6b0dd334293237a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7ba11e209131cea89728.md>)

Recorded evaluation configuration: {"Backbone":"Distill-Qwen-32B","Strategy":"InftyThink"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.
