Benchmarks · speed
Latency (Time to First Token)
Seconds from request to first token as measured by AA, averaged across its workloads over the past 5 days. It may be a reasoning token, so this is not time to first answer.
Artificial Analysis changed how it measures speed in Performance Benchmarking v3.0.0. Values before Nov 4, 2026 are not comparable with later ones. Their method
162 models · 298 results · data as of 2026-11-06 · artificialanalysis.ai ↗
Leaderboard
Gemini 2.5 Flash Lite none
All results
298 rows| # | Model | Effort | Score |
|---|---|---|---|
| 1 | Gemini 2.5 Flash Lite | none | 0.3s |
| 2 | Mistral | none | 0.5s |
| 3 | Llama 3 70b | none | 0.6s |
| 4 | Gemini 2.5 Flash | none | 0.6s |
| 5 | Llama 3 8b | none | 0.6s |
| 6 | Qwen 3 30B-A3B | 0.6s | |
| 7 | Mixtral 8x7B Instruct | none | 0.7s |
| 8 | Mistral Small 3 | none | 0.7s |
| 9 | Qwen 3 14B | 0.7s | |
| 10 | GPT-5.4 Mini | none | 0.7s |
| 11 | MiMo V2.5 Pro | 0.8s | |
| 12 | GPT-5.4 Nano | none | 0.8s |
| 13 | Qwen 3 32B | 0.8s | |
| 14 | GPT-5.6 Luna | none | 0.8s |
| 15 | DeepSeek V3 | none | 0.8s |
| 16 | Gemma 4 31B | 0.9s | |
| 17 | Nemotron 3.5 Lightning | 0.9s | |
| 18 | Gemini 3.5 Flash | minimal | 1.0s |
| 19 | GPT-5 Nano | minimal | 1.0s |
| 20 | GPT-4.1 Nano | none | 1.0s |
| 21 | Qwen 3.5 9B | none | 1.0s |
| 22 | GPT-6 Luna | none | 1.1s |
| 23 | GPT-5 Mini | minimal | 1.1s |
| 24 | DeepSeek V4.1 Flash | max | 1.2s |
| 25 | gpt-oss-20b | high | 1.2s |
| 26 | GPT-6 Sol | none | 1.2s |
| 27 | GPT-5.6 Terra | none | 1.2s |
| 28 | GPT-5.5 | none | 1.3s |
| 29 | DeepSeek V4.1 Flash 1017 | high | 1.3s |
| 30 | Llama 4 Scout | none | 1.3s |