Agentic Reasoning on PinchBench
91.3ScoreDS-v4-Flash
Evaluation Results
| Method | Links | |
|---|---|---|
| DS-v4-FlashParameter Count=284B-A13B2026.06 | 91.3 | |
| Kimi-K2.6Parameter Count=1T-A32B2026.06 | 90.2 | |
| Nemotron 3 UltraParameter Count=550B-A55B2026.06 | 90 | |
| DS-v4-ProParameter Count=1.6T-A49B2026.06 | 88.6 | |
| Qwen-3.5Parameter Count=397B-17B2026.06 | 86.6 | |
| Kimi-K2.5Mode=Instant2026.06 | 85.48 | |
| DeepSeek-V3.2Thinking Mode=nothink2026.06 | 85.38 | |
| Ling-2.6-1T2026.06 | 85.24 | |
| GLM-5Thinking Mode=non-thinking2026.06 | 83.29 | |
| Ling-2.6-flash2026.06 | 81.3 | |
| GLM-5.1Parameter Count=744B-A40B2026.06 | 81.2 | |
| MiniMax-2.7Parameter Count=230B-A10B2026.06 | 77.6 | |
| GPT-5.4Reasoning Mode=non-reasoning2026.06 | 73.4 | |
| Nemotron-3-Super 120B-A12BEvaluation Setting=non-reasoning2026.06 | 73.1 | |
| GPT-5.4-miniEvaluation Setting=non-reasoning2026.06 | 71.4 | |
| GPT-OSS-120BEvaluation Setting=low2026.06 | 52 |