DeepSeek V4 Flash 0731 (free)
DeepSeek🇨🇳
deepseek/deepseek-v4-flash-0731:freeDeepSeek V4 Flash 0731 is a sparse mixture-of-experts model from DeepSeek, with 13B active parameters out of 284B total.
Cost rate
Free
Context
1M
Released
Jul 31, 2026
Input
Text
Output
Text
Support
ReasoningTool callingStructured outputs
Best at
The categories where this model ranks highest.
Life, Physical, & Social Science
#83 · top 21%
Legal & Government
#86 · top 23%
Writing, Literature, & Language
#87 · top 22%
Entertainment, Sports, & Media
#87 · top 22%
Performance
Median latency and throughput measured across recent requests.
Throughput212 tok/s
Latency