STEM / Multimodal Reasoning on MathVista (Accuracy)
83.97AccuracySTEP3-VL-10B
Evaluation results
| Method | Links | |
|---|---|---|
| STEP3-VL-10BNumber of Parameters=10B2026.01 | 83.97 | |
| GLM-4.6V FlashNumber of Parameters=9B2026.01 | 82.85 | |
| MiMo-VL RL-2508Number of Parameters=7B2026.01 | 79.86 | |
| Qwen3-VL ThinkingNumber of Parameters=8B2026.01 | 78.5 | |
| InternVL 3.5Number of Parameters=8B2026.01 | 76.78 | |
| CodePercept-32B-S1LLM Solver=Qwen3-235A22-Thinking2026.03 | 75.9 | |
| Gemini2.5-ProLLM Solver=Qwen3-30A3-Thinking2026.03 | 74.8 | |
| Qwen3-VL-32B-InstructLLM Solver=Qwen3-235A22-Thinking2026.03 | 74 | |
| Qwen3-VL-235A22B-InstructLLM Solver=Qwen3-30A3-Thinking2026.03 | 73.8 | |
| CodePercept-32B-S1LLM Solver=Qwen3-30A3-Thinking2026.03 | 72.9 | |
| Qwen3-VL-32B-InstructLLM Solver=Qwen3-30A3-Thinking2026.03 | 72.2 | |
| CodePercept-8B-S1LLM Solver=Qwen3-235A22-Thinking2026.03 | 71.4 | |
| CodePercept-4B-S1LLM Solver=Qwen3-235A22-Thinking2026.03 | 71.3 | |
| Claude-Opus 4.1-ThinkingLLM Solver=Qwen3-30A3-Thinking2026.03 | 71.1 | |
| Qwen3-VL-8B-InstructLLM Solver=Qwen3-235A22-Thinking2026.03 | 71 | |
| CodePercept-8B-S1LLM Solver=Qwen3-30A3-Thinking2026.03 | 70.2 | |
| CodePercept-4B-S1LLM Solver=Qwen3-30A3-Thinking2026.03 | 69.6 | |
| Qwen3-VL-8B-InstructLLM Solver=Qwen3-30A3-Thinking2026.03 | 69.6 | |
| Qwen3-VL-4B-InstructLLM Solver=Qwen3-235A22-Thinking2026.03 | 69.2 | |
| Qwen3-VL-30A3B-InstructLLM Solver=Qwen3-30A3-Thinking2026.03 | 68 | |
| InternVL3.5-8BLLM Solver=Qwen3-30A3-Thinking2026.03 | 67.7 | |
| Qwen2.5-VL-72BLLM Solver=Qwen3-30A3-Thinking2026.03 | 67.5 | |
| Qwen3-VL-4B-InstructLLM Solver=Qwen3-30A3-Thinking2026.03 | 67.3 | |
| MiniCPM-V-4.5LLM Solver=Qwen3-30A3-Thinking2026.03 | 66.6 | |
| Intern-S1-8BLLM Solver=Qwen3-30A3-Thinking2026.03 | 65.7 | |
| GPT5-ThinkingLLM Solver=Qwen3-30A3-Thinking2026.03 | 65.2 | |
| KeyeVL1.5-8BLLM Solver=Qwen3-30A3-Thinking2026.03 | 64.9 | |
| GLM-4.1V-9BLLM Solver=Qwen3-30A3-Thinking2026.03 | 64.6 |