Measured over the last 30 days
- Models Tracked
- 32
- Avg Tokens / Second
- 9.47
- Avg Time to First Token (ms)
- 14397.50
- Last Updated
- Oct 17, 2026
All siliconflow models
32 live · one row per model · fastest first · 1 folded behind their current lane| Model | Avg Toks/Sec | Min | Max | Avg TTF (ms) |
|---|---|---|---|---|
| Qwen3 VL 30B A3B Instruct | 28.90 | 28.90 | 28.90 | 1170.00 |
| Gemma 4 26B A4B | 27.00 | 19.50 | 35.70 | 1740.00 |
| Gemma 4 31B | 21.00 | 15.40 | 26.50 | 1770.00 |
| Qwen3 30B A3B Instruct 2507 | 19.20 | 17.20 | 21.20 | 1210.00 |
| Qwen3 Coder 30B A3B Instruct | 19.10 | 19.10 | 19.10 | 1170.00 |
| DeepSeek V3 0324 | 18.50 | 18.50 | 18.50 | 1440.00 |
| DeepSeek V3.1 | 14.20 | 14.20 | 14.20 | 1910.00 |
| GLM 5.2 | 13.90 | 2.77 | 23.80 | 8360.00 |
| DeepSeek V4 Flash 0731 | 13.10 | 12.60 | 13.50 | 4030.00 |
| DeepSeek V3.1 Terminus | 12.90 | 12.90 | 12.90 | 1590.00 |
| DeepSeek V4 Flash 0423 | 12.30 | 4.23 | 21.70 | 6140.00 |
| Step 3.5 Flash | 11.10 | 11.10 | 11.10 | 4970.00 |
| Kimi K2.7 Code | 10.10 | 3.13 | 19.10 | 7060.00 |
| gpt-oss-20b | 9.77 | 5.35 | 18.90 | 7080.00 |
| DeepSeek V3.2 | 9.05 | 4.62 | 11.80 | 2750.00 |
| DeepSeek V3.2 Exp | 8.11 | 2.62 | 13.60 | 1730.00 |
| gpt-oss-120b | 6.25 | 2.08 | 10.40 | 13700.00 |
| DeepSeek V4 Pro 0423 | 5.29 | 2.27 | 9.20 | 14400.00 |
| Qwen3 VL 30B A3B Thinking | 4.58 | 4.58 | 4.58 | 12400.00 |
| Qwen3.5-35B-A3B | 4.51 | 4.51 | 4.51 | 13600.00 |
| GLM 5 | 4.47 | 4.47 | 4.47 | 13500.00 |
| DeepSeek V4.1 Flash | 4.31 | 3.05 | 5.21 | 15100.00 |
| Hunyuan A13B Instruct | 3.92 | 3.92 | 3.92 | 14900.00 |
| Qwen3.5-122B-A10B | 3.88 | 3.88 | 3.88 | 15700.00 |
| Qwen3.6 35B A3B | 3.56 | 1.27 | 5.88 | 23400.00 |
| Kimi K2.5 | 3.04 | 3.04 | 3.04 | 17800.00 |
| Kimi K2.6 | 2.60 | 1.10 | 4.71 | 30100.00 |
| GLM 5.1 | 2.22 | 1.09 | 3.00 | 32500.00 |
| R1 0528 | 2.11 | 2.10 | 2.12 | 27100.00 |
| Qwen3.5-27B | 1.82 | 1.82 | 1.82 | 33600.00 |
| Qwen3.5-9B | 1.68 | 1.68 | 1.68 | 36300.00 |
| Qwen3 32B | 0.67 | 0.67 | 0.67 | 92500.00 |
Frequently Asked Questions
Which siliconflow model is fastest?
Based on recent tests, Qwen3 VL 30B A3B Instruct shows the highest average throughput among tracked siliconflow models.
How many recent measurements feed this dashboard?
This provider summary aggregates 2925 individual prompts measured across 2566 monitoring runs over the past month.