Measured over the last 30 days
- Models Tracked
- 30
- Avg Tokens / Second
- 12.08
- Avg Time to First Token (ms)
- 14513.00
- Last Updated
- Oct 17, 2026
All venice models
30 live · one row per model · fastest first · 3 folded behind their current lane| Model | Avg Toks/Sec | Min | Max | Avg TTF (ms) |
|---|---|---|---|---|
| Qwen3 Coder 480B A35B | 44.20 | 44.20 | 44.20 | 730.00 |
| Mistral Small 3.2 24B | 32.00 | 12.50 | 51.50 | 510.00 |
| Uncensored | 28.10 | 28.10 | 28.10 | 630.00 |
| GLM 5.2 | 23.90 | 16.30 | 31.60 | 2250.00 |
| Gemma 4 31B | 23.50 | 21.80 | 24.40 | 610.00 |
| Gemma 4 26B A4B | 20.90 | 3.60 | 37.60 | 1400.00 |
| Kimi K2.7 Code | 18.10 | 11.10 | 30.20 | 3000.00 |
| Qwen3 VL 235B A22B Instruct | 15.80 | 15.80 | 15.80 | 820.00 |
| Qwen3.8 2.4T A95B | 15.50 | 15.50 | 15.50 | 3110.00 |
| MiMo-V2.5 | 15.10 | 15.10 | 15.10 | 2560.00 |
| MiMo-V2.6-Flash | 13.40 | 13.40 | 13.40 | 2750.00 |
| MiniMax M3 | 13.10 | 4.31 | 19.60 | 5410.00 |
| Qwen3 235B A22B Instruct 2507 | 12.40 | 6.39 | 21.20 | 1690.00 |
| DeepSeek V4 Flash 0731 | 10.70 | 10.70 | 10.70 | 3730.00 |
| Nemotron 3.5 Super VL | 9.37 | 9.37 | 9.37 | 6830.00 |
| Qwen3.5-35B-A3B | 9.12 | 9.12 | 9.12 | 6770.00 |
| DeepSeek V4 Pro 0423 | 8.34 | 2.19 | 22.70 | 11700.00 |
| DeepSeek V3.2 | 7.85 | 2.21 | 10.80 | 2620.00 |
| Qwen3.6 35B A3B | 7.14 | 6.16 | 8.82 | 8940.00 |
| Qwen3.5-9B | 6.96 | 6.96 | 6.96 | 8830.00 |
| DeepSeek V4 Flash 0423 | 4.85 | 2.15 | 8.74 | 14300.00 |
| Kimi K2.5 | 4.62 | 3.85 | 5.40 | 12600.00 |
| Qwen3.5 397B A17B | 3.84 | 2.61 | 6.79 | 17800.00 |
| MiniMax M2.5 | 3.70 | 0.68 | 7.84 | 33000.00 |
| Kimi K2.6 | 2.28 | 0.64 | 3.24 | 45400.00 |
| gpt-oss-120b | 2.24 | 2.24 | 2.24 | 21700.00 |
| GLM 4.7 Flash | 1.81 | 1.81 | 1.81 | 23700.00 |
| GLM 5.1 | 1.58 | 1.58 | 1.58 | 38500.00 |
| Qwen3 235B A22B Thinking 2507 | 1.26 | 0.72 | 1.80 | 60300.00 |
| GLM 4.6 | 0.66 | 0.66 | 0.66 | 93200.00 |
Frequently Asked Questions
Which venice model is fastest?
Based on recent tests, Qwen3 Coder 480B A35B shows the highest average throughput among tracked venice models.
How many recent measurements feed this dashboard?
This provider summary aggregates 3237 individual prompts measured across 2813 monitoring runs over the past month.