↑ Higher is bettertok/s287 models
Speed Benchmark
The Speed Benchmark measures output token generation throughput in tokens per second. Speed is measured under standard conditions and reflects the model's raw generation efficiency. Higher throughput generally means faster responses for chat and generation-heavy workloads.
Leading models
Top 8 ranked models by speed · tok/s
1?Mercury 2
1121.62?LFM2.5-VL-1.6B
421.43?Step 3.7 Flash
409.64?Granite 3.3 8B
385.45?Granite 4.0 H Small
383.76NNVIDIA Nemotron 3 Super
363.77?HyperNova 60B 2605
352.98GGemini 3.1 Flash-Lite
339.0287 models with data263 without data
| Rank | Model | Provider | Type | Speed(tok/s) |
|---|---|---|---|---|
| 1 | Mercury 2 | ?Inception | Open weights | 1121.6 |
| 2 | LFM2.5-VL-1.6B | ?Liquid AI | Open weights | 421.4 |
| 3 | Step 3.7 Flash | ?StepFun | Open weights | 409.6 |
| 4 | Granite 3.3 8B | ?IBM | Open weights | 385.4 |
| 5 | Granite 4.0 H Small | ?IBM | Open weights | 383.7 |
| 6 | NVIDIA Nemotron 3 Super | NNVIDIA | Open weights | 363.7 |
| 7 | HyperNova 60B 2605 | ?Multiverse Computing | Open weights | 352.9 |
| 8 | Gemini 3.1 Flash-Lite | GGoogle | Open weights | 339.0 |
| 9 | Nova Micro | ?Amazon | Open weights | 331.6 |
| 10 | gpt-oss-120b (low) | AIOpenAI | Open weights | 328.2 |
| 11 | LFM2.5-8B-A1B | ?Liquid AI | Open weights | 323.9 |
| 12 | gpt-oss-120b (high) | AIOpenAI | Open weights | 314.1 |
| 13 | Llama 3.1 Nemotron 70B | MNVIDIA | Open weights | 301.5 |
| 14 | Gemini 2.5 Flash-Lite | GGoogle | Open weights | 287.0 |
| 15 | NVIDIA Nemotron Nano 12B v2 VL | NNVIDIA | Open weights | 282.6 |
| 16 | Nemotron 3 Nano Omni 30B A3B Reasoning | NNVIDIA | Open weights | 279.0 |
| 17 | Gemini 2.5 Flash-Lite | GGoogle | Open weights | 264.1 |
| 18 | Qwen3.5 Omni Flash | QAlibaba | Open weights | 262.8 |
| 19 | Nemotron 3 Ultra | NNVIDIA | Open weights | 250.2 |
| 20 | NVIDIA Nemotron Nano 12B v2 VL | NNVIDIA | Open weights | 241.4 |
| 21 | gpt-oss-20b (low) | AIOpenAI | Open weights | 228.4 |
| 22 | Gemini 2.5 Flash | GGoogle | Open weights | 226.5 |
| 23 | o3-mini | AIOpenAI | Open weights | 226.1 |
| 24 | Gemini 2.5 Flash | GGoogle | Open weights | 222.4 |
| 25 | GLM-5.2 (max) | ?Z AI | Open weights | 217.6 |
| 26 | o3-mini (high) | AIOpenAI | Open weights | 217.2 |
| 27 | GPT-4o (Nov) | AIOpenAI | Open weights | 206.8 |
| 28 | Qwen3.7 Max | QAlibaba | Open weights | 206.4 |
| 29 | Trinity Large Thinking | ?Arcee AI | Open weights | 206.3 |
| 30 | Hy3-preview | ?Tencent | Open weights | 202.9 |
| 31 | GPT-5.1 Codex mini (high) | AIOpenAI | Open weights | 200.7 |
| 32 | GPT-5.1 Codex (high) | AIOpenAI | Open weights | 200.4 |
| 33 | Hy3-preview | ?Tencent | Open weights | 196.2 |
| 34 | gpt-oss-20b (high) | AIOpenAI | Open weights | 195.3 |
| 35 | Qwen3.5 35B A3B | QAlibaba | Open weights | 195.2 |
| 36 | GPT-5 (ChatGPT) | AIOpenAI | Open weights | 195.0 |
| 37 | Qwen3 Next 80B A3B | QAlibaba | Open weights | 193.0 |
| 38 | Gemini 3.5 Flash | GGoogle | Open weights | 192.2 |
| 39 | Gemini 3.5 Flash (minimal) | GGoogle | Open weights | 191.6 |
| 40 | Gemini 3.5 Flash (medium) | GGoogle | Open weights | 191.4 |
| 41 | Command A+ | ?Cohere | Open weights | 190.8 |
| 42 | GPT-5 Codex (high) | AIOpenAI | Open weights | 190.6 |
| 43 | GPT-4.1 nano | AIOpenAI | Open weights | 188.4 |
| 44 | GPT-5.4 mini (medium) | AIOpenAI | Open weights | 184.3 |
| 45 | GPT-5.4 mini | AIOpenAI | Open weights | 184.1 |
| 46 | Step 3.5 Flash 2603 | ?StepFun | Open weights | 181.8 |
| 47 | Step 3.5 Flash | ?StepFun | Open weights | 181.7 |
| 48 | Jamba 1.6 Mini | ?AI21 Labs | Open weights | 180.2 |
| 49 | Qwen3.6 35B A3B | QAlibaba | Open weights | 179.9 |
| 50 | GPT-5.4 mini (xhigh) | AIOpenAI | Open weights | 179.3 |
Page 1 of 11 · 550 results