# STEM & Reasoning on MathVerse Source: [STEM & Reasoning on MathVerse]() Collection index: [llms.txt]() Data updated: 2026-03-12T09:09:05.013Z Evaluation of model performance on the MathVerse dataset covering STEM subjects and general reasoning capabilities, measured by 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 | STEM & Reasoning | | Dataset (including any stated split) | MathVerse | | Metric | Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 28 of 28 | | 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](): Higher is better; unit: PERCENT. ## Recorded results | Method | Accuracy | Result date | Report paper | | --- | --- | --- | --- | | Gemini 3-Pro | 91.62 | Not recorded | 69842436a8b01e76cc8ee882 | | Gemini 2.5-Pro | 86.27 | Not recorded | 69842436a8b01e76cc8ee882 | | ERNIE 5.0 | 85.13 | Not recorded | 69842436a8b01e76cc8ee882 | | GPT-5 | 84.19 | Not recorded | 69842436a8b01e76cc8ee882 | | Qwen3-VL | 83.96 | Not recorded | 69842436a8b01e76cc8ee882 | | CodePercept-32B-S1 | 73.56 | Not recorded | 69b248b5d030084c22e2fafd | | Gemini2.5-Pro | 73.47 | Not recorded | 69b248b5d030084c22e2fafd | | CodePercept-32B-S1 | 71.7 | Not recorded | 69b248b5d030084c22e2fafd | | Qwen3-VL-32B-Instruct | 71.09 | Not recorded | 69b248b5d030084c22e2fafd | | Qwen3-VL-235A22B-Instruct | 70.08 | Not recorded | 69b248b5d030084c22e2fafd | | Qwen3-VL-32B-Instruct | 69.9 | Not recorded | 69b248b5d030084c22e2fafd | | GPT5-Thinking | 69.56 | Not recorded | 69b248b5d030084c22e2fafd | | CodePercept-8B-S1 | 67.95 | Not recorded | 69b248b5d030084c22e2fafd | | CodePercept-4B-S1 | 66.73 | Not recorded | 69b248b5d030084c22e2fafd | | CodePercept-8B-S1 | 66.52 | Not recorded | 69b248b5d030084c22e2fafd | | Qwen3-VL-30A3B-Instruct | 66.44 | Not recorded | 69b248b5d030084c22e2fafd | | Qwen3-VL-4B-Instruct | 66.39 | Not recorded | 69b248b5d030084c22e2fafd | | CodePercept-4B-S1 | 65.59 | Not recorded | 69b248b5d030084c22e2fafd | | Qwen3-VL-4B-Instruct | 64.59 | Not recorded | 69b248b5d030084c22e2fafd | | Qwen3-VL-8B-Instruct | 63.88 | Not recorded | 69b248b5d030084c22e2fafd | | Qwen3-VL-8B-Instruct | 63.75 | Not recorded | 69b248b5d030084c22e2fafd | | MiniCPM-V-4.5 | 57.84 | Not recorded | 69b248b5d030084c22e2fafd | | Claude-Opus 4.1-Thinking | 56.19 | Not recorded | 69b248b5d030084c22e2fafd | | Qwen2.5-VL-72B | 55.4 | Not recorded | 69b248b5d030084c22e2fafd | | GLM-4.1V-9B | 54.47 | Not recorded | 69b248b5d030084c22e2fafd | | InternVL3.5-8B | 53.4 | Not recorded | 69b248b5d030084c22e2fafd | | Intern-S1-8B | 51.9 | Not recorded | 69b248b5d030084c22e2fafd | | KeyeVL1.5-8B | 49.95 | Not recorded | 69b248b5d030084c22e2fafd | ### Gemini 3-Pro Accuracy: 91.62. Result ID: 69964b4b5698665ec48f963a. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Gemini 2.5-Pro Accuracy: 86.27. Result ID: 69964b4b5698665ec48f9638. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### ERNIE 5.0 Accuracy: 85.13. Result ID: 69964b4b5698665ec48f963b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### GPT-5 Accuracy: 84.19. Result ID: 69964b4b5698665ec48f9639. [Reporting paper]() Recorded evaluation configuration: {"tier":"High"}. This record does not assert independent reproduction. ### Qwen3-VL Accuracy: 83.96. Result ID: 69964b4b5698665ec48f9637. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"mode":"Thinking"}. This record does not assert independent reproduction. ### CodePercept-32B-S1 Accuracy: 73.56. Result ID: 69b282956dbb7075b95d5af2. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"LLM Solver":"Qwen3-235A22-Thinking"}. This record does not assert independent reproduction. ### Gemini2.5-Pro Accuracy: 73.47. Result ID: 69b282956dbb7075b95d5ae6. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"LLM Solver":"Qwen3-30A3-Thinking"}. This record does not assert independent reproduction. ### CodePercept-32B-S1 Accuracy: 71.7. Result ID: 69b282956dbb7075b95d5aec. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"LLM Solver":"Qwen3-30A3-Thinking"}. This record does not assert independent reproduction. ### Qwen3-VL-32B-Instruct Accuracy: 71.09. Result ID: 69b282956dbb7075b95d5aeb. [Reporting paper]() Recorded evaluation configuration: {"LLM Solver":"Qwen3-30A3-Thinking"}. This record does not assert independent reproduction. ### Qwen3-VL-235A22B-Instruct Accuracy: 70.08. Result ID: 69b282956dbb7075b95d5ae5. [Reporting paper]() Recorded evaluation configuration: {"LLM Solver":"Qwen3-30A3-Thinking"}. This record does not assert independent reproduction. ### Qwen3-VL-32B-Instruct Accuracy: 69.9. Result ID: 69b282956dbb7075b95d5af1. [Reporting paper]() Recorded evaluation configuration: {"LLM Solver":"Qwen3-235A22-Thinking"}. This record does not assert independent reproduction. ### GPT5-Thinking Accuracy: 69.56. Result ID: 69b282956dbb7075b95d5ae4. [Reporting paper]() Recorded evaluation configuration: {"LLM Solver":"Qwen3-30A3-Thinking"}. This record does not assert independent reproduction. ### CodePercept-8B-S1 Accuracy: 67.95. Result ID: 69b282956dbb7075b95d5af0. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"LLM Solver":"Qwen3-235A22-Thinking"}. This record does not assert independent reproduction. ### CodePercept-4B-S1 Accuracy: 66.73. Result ID: 69b282956dbb7075b95d5aee. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"LLM Solver":"Qwen3-235A22-Thinking"}. This record does not assert independent reproduction. ### CodePercept-8B-S1 Accuracy: 66.52. Result ID: 69b282956dbb7075b95d5aea. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"LLM Solver":"Qwen3-30A3-Thinking"}. This record does not assert independent reproduction. ### Qwen3-VL-30A3B-Instruct Accuracy: 66.44. Result ID: 69b282956dbb7075b95d5ae2. [Reporting paper]() Recorded evaluation configuration: {"LLM Solver":"Qwen3-30A3-Thinking"}. This record does not assert independent reproduction. ### Qwen3-VL-4B-Instruct Accuracy: 66.39. Result ID: 69b282956dbb7075b95d5aed. [Reporting paper]() Recorded evaluation configuration: {"LLM Solver":"Qwen3-235A22-Thinking"}. This record does not assert independent reproduction. ### CodePercept-4B-S1 Accuracy: 65.59. Result ID: 69b282956dbb7075b95d5ae8. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"LLM Solver":"Qwen3-30A3-Thinking"}. This record does not assert independent reproduction. ### Qwen3-VL-4B-Instruct Accuracy: 64.59. Result ID: 69b282956dbb7075b95d5ae7. [Reporting paper]() Recorded evaluation configuration: {"LLM Solver":"Qwen3-30A3-Thinking"}. This record does not assert independent reproduction. ### Qwen3-VL-8B-Instruct Accuracy: 63.88. Result ID: 69b282956dbb7075b95d5aef. [Reporting paper]() Recorded evaluation configuration: {"LLM Solver":"Qwen3-235A22-Thinking"}. This record does not assert independent reproduction. ### Qwen3-VL-8B-Instruct Accuracy: 63.75. Result ID: 69b282956dbb7075b95d5ae9. [Reporting paper]() Recorded evaluation configuration: {"LLM Solver":"Qwen3-30A3-Thinking"}. This record does not assert independent reproduction. ### MiniCPM-V-4.5 Accuracy: 57.84. Result ID: 69b282956dbb7075b95d5ae0. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"LLM Solver":"Qwen3-30A3-Thinking"}. This record does not assert independent reproduction. ### Claude-Opus 4.1-Thinking Accuracy: 56.19. Result ID: 69b282956dbb7075b95d5ae3. [Reporting paper]() Recorded evaluation configuration: {"LLM Solver":"Qwen3-30A3-Thinking"}. This record does not assert independent reproduction. ### Qwen2.5-VL-72B Accuracy: 55.4. Result ID: 69b282956dbb7075b95d5ae1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"LLM Solver":"Qwen3-30A3-Thinking"}. This record does not assert independent reproduction. ### GLM-4.1V-9B Accuracy: 54.47. Result ID: 69b282956dbb7075b95d5ade. [Reporting paper]() Recorded evaluation configuration: {"LLM Solver":"Qwen3-30A3-Thinking"}. This record does not assert independent reproduction. ### InternVL3.5-8B Accuracy: 53.4. Result ID: 69b282956dbb7075b95d5adf. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"LLM Solver":"Qwen3-30A3-Thinking"}. This record does not assert independent reproduction. ### Intern-S1-8B Accuracy: 51.9. Result ID: 69b282956dbb7075b95d5add. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"LLM Solver":"Qwen3-30A3-Thinking"}. This record does not assert independent reproduction. ### KeyeVL1.5-8B Accuracy: 49.95. Result ID: 69b282956dbb7075b95d5adc. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"LLM Solver":"Qwen3-30A3-Thinking"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.