Home / Best AI for… / Fastest AI Models
Task ranking
Fastest AI Models
Ranked by measured output speed (tokens per second). Only models with a sourced speed measurement appear — coverage is stated, never padded.
Our top pick: DeepSeek V4.1 Flash — 213 tok/s measured output speed · League Score 75 · $0.30/$1.20 per 1M. Full record, pricing and benchmarks on its model page; or run the model picker for a recommendation matched to your budget and context needs.
| # | Model | Maker | Why it ranks here |
|---|---|---|---|
| 1 | DeepSeek V4.1 Flash | DeepSeek | 213 tok/s measured output speed · League Score 75 · $0.30/$1.20 per 1M |
| 2 | Grok 4.3 | xAI (SpaceXAI) | 151.8 tok/s measured output speed · League Score 74 · $1.25/$2.50 per 1M |
| 3 | DeepSeek V4 Pro (0813) | DeepSeek | 99.4 tok/s measured output speed · League Score 63 · $1.32/$3.96 per 1M |
| 4 | Grok 4.7 | xAI (SpaceXAI) | 87.6 tok/s measured output speed · League Score 67 · $2.00/$6.00 per 1M |
| 5 | Claude Opus 5.5 | Anthropic | 75.6 tok/s measured output speed · League Score 96 · $4.00/$20.00 per 1M |
| 6 | Kimi K3 | Moonshot AI | 44.8 tok/s measured output speed · League Score 63 · $3.00/$15.00 per 1M |
| 7 | Qwen3.8-Max | Alibaba Qwen | 37 tok/s measured output speed · League Score 62 · $2.00/$6.00 per 1M |
Coverage: 7 of 53 tracked models have a sourced speed measurement (Artificial Analysis, cited on each model page). Unmeasured models are excluded, not estimated.
Rankings regenerate from the model database on every data update. Metrics and sources: methodology.