Provider Snapshot
Models Tracked
77
Avg Tokens / Second
35.40
Avg Time to First Token (ms)
0.00
Last Updated
Jun 10, 2026
Key Takeaways
77 deepinfra models are actively benchmarked with 2972 total measurements across 2582 benchmark runs.
qwen-3.5-2b leads the fleet with 189.00 tokens/second, while GPT-oss-120b-Turbo delivers 80.10 tok/s.
Performance varies by 136.0% across the deepinfra model lineup, indicating diverse optimization strategies for different use cases.
The deepinfra model fleet shows varied performance characteristics (85.0% variation coefficient), reflecting diverse model architectures.
Fastest Models
| Provider | Model | Avg Toks/Sec | Min | Max | Avg TTF (ms) |
|---|---|---|---|---|---|
| deepinfra | qwen-3.5-2b | 189.00 | 6.88 | 242.00 | 0.00 |
| deepinfra | Qwen3.6-35B-A3B | 117.00 | 70.80 | 160.00 | 0.00 |
| deepinfra | qwen-3.5-35b-a3b | 116.00 | 20.50 | 161.00 | 0.00 |
| deepinfra | Nemotron-3-Nano-Omni-30B-A3B-Reasoning | 113.00 | 56.40 | 180.00 | 0.00 |
| deepinfra | MiMo-V2.5 | 92.20 | 42.20 | 138.00 | 0.00 |
| deepinfra | GPT-oss-120b-Turbo | 80.10 | 23.10 | 184.00 | 0.00 |
All Models
Complete list of all deepinfra models tracked in the benchmark system. Click any model name to view detailed performance data.
| Provider | Model | Avg Toks/Sec | Min | Max | Avg TTF (ms) |
|---|---|---|---|---|---|
| deepinfra | Seed-2.0-code | 0.45 | 0.45 | 0.45 | 0.00 |
| deepinfra | Seed-2.0-pro | 2.08 | 0.93 | 2.85 | 0.00 |
| deepinfra | MiniMax-M2.5 | 14.20 | 4.76 | 27.00 | 0.00 |
| deepinfra | MiniMax-M2.7 | 11.40 | 4.82 | 24.30 | 0.00 |
| deepinfra | qwen-2.5-72b | 18.60 | 1.61 | 35.70 | 0.00 |
| deepinfra | Qwen 2.5 Coder 32B | 47.90 | 14.60 | 70.80 | 0.00 |
| deepinfra | qwen-3-14b | 32.60 | 1.64 | 59.10 | 0.00 |
| deepinfra | qwen-3-235b | 14.70 | 2.47 | 46.90 | 0.00 |
| deepinfra | Qwen3-30B-A3B | 49.50 | 18.60 | 70.60 | 0.00 |
| deepinfra | Qwen3-32B | 43.30 | 14.10 | 65.80 | 0.00 |
| deepinfra | Qwen3-Coder-480B-A35B-Instruct-Turbo | 14.30 | 4.28 | 22.90 | 0.00 |
| deepinfra | Qwen3-Max | 19.90 | 13.40 | 26.40 | 0.00 |
| deepinfra | Qwen3-Max-Thinking | 17.10 | 12.60 | 20.40 | 0.00 |
| deepinfra | Qwen3-Next-80B-A3B-Instruct | 49.20 | 2.65 | 86.30 | 0.00 |
| deepinfra | qwen-3.5-0.8b | 76.00 | 24.70 | 207.00 | 0.00 |
| deepinfra | qwen-3.5-122b-a10b | 79.40 | 31.20 | 115.00 | 0.00 |
| deepinfra | qwen-3.5-27b | 35.50 | 4.26 | 73.40 | 0.00 |
| deepinfra | qwen-3.5-2b | 189.00 | 6.88 | 242.00 | 0.00 |
| deepinfra | qwen-3.5-35b-a3b | 116.00 | 20.50 | 161.00 | 0.00 |
| deepinfra | qwen-3.5-397b-a17b | 63.20 | 20.00 | 102.00 | 0.00 |
| deepinfra | Qwen3.6-35B-A3B | 117.00 | 70.80 | 160.00 | 0.00 |
| deepinfra | Qwen3.7-Max | 2.18 | 1.48 | 3.97 | 0.00 |
| deepinfra | MiMo-V2.5 | 92.20 | 42.20 | 138.00 | 0.00 |
| deepinfra | MiMo-V2.5-Pro | 38.20 | 2.86 | 69.40 | 0.00 |
| deepinfra | claude-haiku-4-5 | 29.30 | 22.00 | 34.10 | 0.00 |
| deepinfra | claude-opus-4-7 | 29.40 | 22.40 | 34.50 | 0.00 |
| deepinfra | claude-opus-4-8 | 26.80 | 20.40 | 30.10 | 0.00 |
| deepinfra | claude-sonnet-4-6 | 15.60 | 9.65 | 18.90 | 0.00 |
| deepinfra | DeepSeek-R1-0528 | 32.10 | 10.80 | 63.80 | 0.00 |
| deepinfra | DeepSeek-V3 | 17.90 | 11.20 | 30.60 | 0.00 |
| deepinfra | DeepSeek-V3.1 | 7.04 | 1.34 | 12.50 | 0.00 |
| deepinfra | DeepSeek-V3.1-Terminus | 20.40 | 9.46 | 44.80 | 0.00 |
| deepinfra | deepseek-v3.2 | 9.73 | 1.02 | 22.90 | 0.00 |
| deepinfra | DeepSeek-V4-Flash | 13.50 | 2.09 | 22.20 | 0.00 |
| deepinfra | DeepSeek-V4-Pro | 24.00 | 1.88 | 47.10 | 0.00 |
| deepinfra | gemini-2.5-flash | 36.00 | 12.10 | 45.70 | 0.00 |
| deepinfra | gemini-3.1-pro | 21.80 | 17.30 | 25.10 | 0.00 |
| deepinfra | gemma-3-12b-it | 24.80 | 11.80 | 37.80 | 0.00 |
| deepinfra | gemma-3-27b-it | 21.10 | 10.30 | 35.30 | 0.00 |
| deepinfra | gemma-4-26B-A4B-it | 28.30 | 9.95 | 46.10 | 0.00 |
| deepinfra | gemma-4-31B-it | 12.50 | 5.07 | 27.10 | 0.00 |
| deepinfra | gemma-4-31B-it-turbo | 17.80 | 2.32 | 46.50 | 0.00 |
| deepinfra | llama-2-70b | 27.10 | 3.08 | 37.00 | 0.00 |
| deepinfra | llama-3.2-11b | 41.20 | 1.06 | 52.50 | 0.00 |
| deepinfra | llama-3.2-1b | 39.70 | 7.79 | 52.60 | 0.00 |
| deepinfra | llama-3.2-3b | 41.20 | 3.15 | 52.70 | 0.00 |
| deepinfra | llama-3.2-90b | 31.50 | 3.47 | 60.00 | 0.00 |
| deepinfra | llama-3.3-70b | 19.20 | 1.49 | 40.10 | 0.00 |
| deepinfra | Llama-3.3-70B-Instruct-Turbo | 14.70 | 1.59 | 27.00 | 0.00 |
| deepinfra | Llama-4-Maverick-17B-128E-Instruct-FP8 | 19.90 | 2.53 | 37.40 | 0.00 |
| deepinfra | Llama-Guard-4-12B | 3.31 | 2.09 | 4.21 | 0.00 |
| deepinfra | llama-3-70b | 29.20 | 3.03 | 37.30 | 0.00 |
| deepinfra | llama-3-8b | 42.60 | 15.70 | 75.60 | 0.00 |
| deepinfra | llama-3.1-405b | 17.40 | 2.63 | 27.40 | 0.00 |
| deepinfra | llama-3.1-70b | 30.60 | 3.56 | 65.90 | 0.00 |
| deepinfra | llama-3.1-8b | 30.10 | 8.79 | 66.30 | 0.00 |
| deepinfra | phi-4 | 55.80 | 53.10 | 58.50 | 0.00 |
| deepinfra | devstral-small | 37.90 | 1.77 | 67.10 | 0.00 |
| deepinfra | mistral-7b | 39.20 | 1.81 | 67.70 | 0.00 |
| deepinfra | Mistral-Nemo-Instruct-2407 | 25.20 | 9.18 | 40.30 | 0.00 |
| deepinfra | Mistral-Small-24B-Instruct-2501 | 44.10 | 6.64 | 56.60 | 0.00 |
| deepinfra | Mistral-Small-3.2-24B-Instruct-2506 | 35.30 | 11.20 | 66.00 | 0.00 |
| deepinfra | Kimi-K2.5 | 8.44 | 3.39 | 27.10 | 0.00 |
| deepinfra | Kimi-K2.6 | 21.60 | 2.40 | 62.80 | 0.00 |
| deepinfra | NVIDIA-Nemotron-3-Super-120B-A12B | 20.00 | 1.73 | 62.50 | 0.00 |
| deepinfra | NVIDIA-Nemotron-3-Ultra-550B-A55B | 46.40 | 11.50 | 73.50 | 0.00 |
| deepinfra | Nemotron-3-Nano-30B-A3B | 52.50 | 9.98 | 74.70 | 0.00 |
| deepinfra | Nemotron-3-Nano-Omni-30B-A3B-Reasoning | 113.00 | 56.40 | 180.00 | 0.00 |
| deepinfra | GPT-oss-120b | 18.60 | 7.51 | 38.60 | 0.00 |
| deepinfra | GPT-oss-120b-Turbo | 80.10 | 23.10 | 184.00 | 0.00 |
| deepinfra | GPT-oss-20b | 18.50 | 1.26 | 53.90 | 0.00 |
| deepinfra | Step-3.5-Flash | 27.40 | 18.70 | 35.60 | 0.00 |
| deepinfra | GLM-4.6 | 34.10 | 9.62 | 67.10 | 0.00 |
| deepinfra | GLM-4.7 | 29.10 | 5.27 | 47.50 | 0.00 |
| deepinfra | GLM-4.7-Flash | 50.60 | 8.91 | 81.30 | 0.00 |
| deepinfra | GLM-5 | 20.30 | 7.07 | 49.20 | 0.00 |
| deepinfra | GLM-5.1 | 27.90 | 5.75 | 47.10 | 0.00 |
Featured Models
Frequently Asked Questions
Based on recent tests, qwen-3.5-2b shows the highest average throughput among tracked deepinfra models.
This provider summary aggregates 2972 individual prompts measured across 2582 monitoring runs over the past month.