Multi-turn Dialogue Evaluation on MT_Bench CodaSet OOD (test)
98.16Performance (%)Qwen3-235B
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Qwen3-235B2026.05 | 98.16 | 3.1 | |
| gpt-oss-120b2026.05 | 98.16 | 2.9 | |
| DecoRBase Model=gpt-oss-120b2026.05 | 98.16 | 2.7 | |
| DeepSeek-V3.12026.05 | 97.24 | 5.3 | |
| DecoRBase Model=DeepSeek-V3.12026.05 | 97.24 | 4.7 | |
| DeepSeek-V3.2-Exp2026.05 | 95.53 | 2.4 | |
| gemma-3-27b-it2026.05 | 95.53 | 1.2 | |
| Random Router2026.05 | 95.26 | 3.7 | |
| RouterDC2026.05 | 95 | 2.4 | |
| EmbedLLM2026.05 | 94.74 | 3.2 | |
| Kimi-K2-Instruct2026.05 | 94.47 | 7.9 | |
| LLM Router2026.05 | 94.34 | 1.7 | |
| MODEL-SAT2026.05 | 94.21 | 2 | |
| KNN Router2026.05 | 93.95 | 1.4 | |
| Mistral-3.2-24B2026.05 | 93.68 | 1.2 | |
| gemma-3-12b-it2026.05 | 92.37 | 1 |