↓ Lower is betters TTFT287 models
Latency Benchmark
The Latency Benchmark measures time to first answer token (TTFT), representing how quickly a model begins generating after receiving a prompt. Lower latency is critical for interactive and real-time applications. Measured values include reasoning prep time where applicable.
Leading models
Top 8 ranked models by latency · s TTFT
1?North Mini Code
0.30s2GGemini 2.5 Flash-Lite
0.32s3?Command A+
0.41s4NNVIDIA Nemotron Nano 12B v2 VL
0.43s5GGemini 2.5 Flash
0.49s6xGrok Build 0.1 0616
0.54s7xGrok 4.20 0309 v2
0.56s8?LFM2 24B A2B
0.57s287 models with data263 without data
| Rank | Model | Provider | Type | Latency(s TTFT) |
|---|---|---|---|---|
| 1 | North Mini Code | ?Cohere | Open weights | 0.30s |
| 2 | Gemini 2.5 Flash-Lite | GGoogle | Open weights | 0.32s |
| 3 | Command A+ | ?Cohere | Open weights | 0.41s |
| 4 | NVIDIA Nemotron Nano 12B v2 VL | NNVIDIA | Open weights | 0.43s |
| 5 | Gemini 2.5 Flash | GGoogle | Open weights | 0.49s |
| 6 | Grok Build 0.1 0616 | xxAI | Open weights | 0.54s |
| 7 | Grok 4.20 0309 v2 | xxAI | Open weights | 0.56s |
| 8 | LFM2 24B A2B | ?Liquid AI | Open weights | 0.57s |
| 9 | NVIDIA Nemotron 3 Nano | NNVIDIA | Open weights | 0.60s |
| 10 | GPT-5.4 nano | AIOpenAI | Open weights | 0.61s |
| 11 | GPT-4.1 nano | AIOpenAI | Open weights | 0.62s |
| 12 | GPT-5.4 mini | AIOpenAI | Open weights | 0.65s |
| 13 | Llama 3.2 11B (Vision) | MMeta | Open weights | 0.68s |
| 14 | Ministral 3 3B | ?Mistral | Open weights | 0.68s |
| 15 | DeepSeek R1 Distill Llama 70B | MDeepSeek | Open weights | 0.71s |
| 16 | GPT-5.4 (Non-reasoning) | AIOpenAI | Open weights | 0.71s |
| 17 | Mistral Small 3.2 | MiMistral | Open weights | 0.72s |
| 18 | Grok 4.3 (Non-reasoning) | xxAI | Open weights | 0.73s |
| 19 | Grok 3 mini Reasoning (high) | xxAI | Open weights | 0.74s |
| 20 | Mistral Small 3.1 | MiMistral | Open weights | 0.74s |
| 21 | Mistral Small (Feb) | MiMistral | Open weights | 0.75s |
| 22 | gpt-oss-20b (high) | AIOpenAI | Open weights | 0.77s |
| 23 | Mistral Small 3 | MiMistral | Open weights | 0.77s |
| 24 | Mistral Small 4 | MiMistral | Open weights | 0.78s |
| 25 | Mistral Small 4 | MiMistral | Open weights | 0.79s |
| 26 | gpt-oss-120b (low) | AIOpenAI | Open weights | 0.80s |
| 27 | Mistral Small (Sep) | MiMistral | Open weights | 0.80s |
| 28 | Qwen3.5 4B | QAlibaba | Open weights | 0.80s |
| 29 | GPT-4o (Nov) | AIOpenAI | Open weights | 0.81s |
| 30 | GPT-5 (ChatGPT) | AIOpenAI | Open weights | 0.81s |
| 31 | Ministral 3 8B | ?Mistral | Open weights | 0.82s |
| 32 | Phi-4 Multimodal | ?Microsoft | Open weights | 0.82s |
| 33 | Mistral 7B | MiMistral | Open weights | 0.83s |
| 34 | GLM-4.7 | ?Z AI | Open weights | 0.84s |
| 35 | gpt-oss-20b (low) | AIOpenAI | Open weights | 0.84s |
| 36 | Granite 4.1 8B | ?IBM | Open weights | 0.84s |
| 37 | Llama 3.2 1B | MMeta | Open weights | 0.84s |
| 38 | Phi-4 Mini | ?Microsoft | Open weights | 0.84s |
| 39 | Cogito v2.1 | ?Deep Cogito | Open weights | 0.85s |
| 40 | Llama 4 Scout | MMeta | Open weights | 0.86s |
| 41 | Ministral 3 14B | ?Mistral | Open weights | 0.86s |
| 42 | gpt-oss-120b (high) | AIOpenAI | Open weights | 0.87s |
| 43 | Gemini 3.5 Flash (minimal) | GGoogle | Open weights | 0.88s |
| 44 | Gemini 3 Flash | GGoogle | Open weights | 0.89s |
| 45 | Claude 4.5 Haiku | AAnthropic | Open weights | 0.92s |
| 46 | GLM-4.7 | ?Z AI | Open weights | 0.92s |
| 47 | GPT-5 mini (minimal) | AIOpenAI | Open weights | 0.92s |
| 48 | Step 3.7 Flash | ?StepFun | Open weights | 0.92s |
| 49 | GPT-4o (May) | AIOpenAI | Open weights | 0.93s |
| 50 | GPT-5 nano (minimal) | AIOpenAI | Open weights | 0.94s |
Page 1 of 11 · 550 results