The fastest Mistral model is Mistral Small 4 at 110 output tokens per second. Mixtral 8x22B Instruct (102 t/s) and Mistral Small 3.2 24B (92 t/s) round out the top three.
AI models ranked by output speed (tokens per second, p50). The fastest large language models for low-latency and high-throughput applications.
The fastest Mistral model is Mistral Small 4 at 110 output tokens per second. Mixtral 8x22B Instruct (102 t/s) and Mistral Small 3.2 24B (92 t/s) round out the top three.
Mixtral 8x22B Instruct (102 t/s) is the closest alternative on this metric, followed by Mistral Small 3.2 24B (92 t/s). See the full ranking above for the tradeoffs.
modelgrep tracks 19 Mistral models with live benchmarks, speed, latency and per-provider pricing, led on intelligence by Mistral Medium 3.5. 16 of them qualify for this ranking.