General Task Performance on Macro-average (Mathematics, Multi-Hop QA, Code Generation)
69AccuracyClaude-Sonnet-4-5
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Claude-Sonnet-4-52026.01 | 69 | 91.4 | |
| Qwen3-235B-A22B-Instruct2026.01 | 67.5 | 88.5 | |
| GPT-4.1-MINI2026.01 | 65.4 | 76.3 | |
| Deepseek-V3.12026.01 | 64.6 | 87.4 | |
| Gemini-2.5-Flash2026.01 | 63.9 | 81.4 | |
| DeepSeek-R1-Distill-Qwen-32B2026.01 | 61.9 | 82.3 | |
| Grok-4-Fast2026.01 | 60.2 | 76.7 | |
| DeepSeek-R1-Distill-Qwen-14B2026.01 | 60 | 79.1 | |
| Qwen2.5-72B-Instruct2026.01 | 59.9 | 78.4 | |
| QwQ-32B2026.01 | 59.1 | 71.5 | |
| Qwen2.5-32B-Instruct2026.01 | 57.5 | 75 | |
| Qwen2.5-14B-Instruct2026.01 | 56.1 | 67.9 | |
| Qwen3-32B2026.01 | 56.1 | 81.1 | |
| Meta-Llama-3.1-70B-Instruct2026.01 | 55.9 | 77.7 | |
| GLM-Z1-32B2026.01 | 54.8 | 64.9 | |
| Qwen2.5-7B-Instruct2026.01 | 50.5 | 65.4 | |
| DeepSeek-R1-Distill-Qwen-7B2026.01 | 48.4 | 75.1 | |
| DeepSeek-R1-Distill-Llama-8B2026.01 | 47.7 | 70.3 | |
| OpenReasoning-Nemotron-14B2026.01 | 45.9 | 72.7 | |
| Meta-Llama-3.1-8B-Instruct2026.01 | 40.5 | 70.8 | |
| Qwen2.5-1.5B-Instruct2026.01 | 27.4 | 64.1 |