Tool Use on Unified evaluation benchmark D Multi-Turn (test)
8.3SP ScoreQwen3-32B
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Qwen3-32BSetting=Hybrid-20 Setting2026.04 | 8.3 | 38.2 | 21.6 | 28.8 | |
| Qwen3-8B (Upper Bound)Setting=GT Setting2026.04 | 0 | 39.5 | 18.8 | 24.9 | |
| GPT-5.2 InstantSetting=Hybrid-20 Setting2026.04 | 0 | 16.1 | 9.1 | 9.8 | |
| Gemini 3 Flash PreviewSetting=Hybrid-20 Setting2026.04 | 0 | 22.5 | 14.2 | 15.6 | |
| Claude 4.6 SonnetSetting=Hybrid-20 Setting2026.04 | 0 | 34.8 | 19.4 | 25.2 | |
| Kimi-K2-InstructSetting=Hybrid-20 Setting2026.04 | 0 | 31.1 | 18.4 | 24.9 | |
| DeepSeek-V3.2Setting=Hybrid-20 Setting2026.04 | 0 | 15 | 10.3 | 13 | |
| Qwen3-8B (Vanilla)Setting=Hybrid-20 Setting2026.04 | 0 | 25.8 | 12.1 | 12.1 | |
| UniToolCallSetting=Hybrid-20 Setting2026.04 | 0 | 39.4 | 21.2 | 26.1 |