Multi-turn Inference Latency on IntentFlow T=37-48
1.2Average Latency (Demand Turns) (ms)Qwen3-30B-A3B
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Qwen3-30B-A3BParameters=30B2026.04 | 1.2 | 1.1 | 1.1 | |
| IntentFlow2026.04 | 1.7 | 1.3 | 1.5 | |
| Gemini-2.5-Flash-Lite2026.04 | 2.2 | 2.1 | 2.2 | |
| DeepSeek-V3.22026.04 | 3.5 | 3.2 | 3.3 | |
| Claude-Haiku-4.52026.04 | 4.1 | 3.4 | 3.7 | |
| Gemini-3-Flash2026.04 | 4.3 | 4.3 | 4.3 | |
| GPT-5-Nano2026.04 | 7.2 | 6.2 | 6.6 | |
| GPT-oss-120bParameters=120b2026.04 | 7.9 | 6.8 | 7.3 | |
| GPT-5-Mini2026.04 | 9.3 | 6.4 | 7.6 | |
| Qwen3.5-Flash2026.04 | 16.4 | 17.8 | 17.2 |