DeepSeek V4.1 Flash Fast

on RunAnywhere
deepseek-v4-1-flash-fast
Released Oct 14, 2026
Context
1.0M
Input
$0.60/M
Output
$2.40/M

The original DeepSeek V4.1 Flash weights served on dedicated fast capacity, at about 2.7x the throughput of the standard endpoint. Built for agent loops. 1M context window, image input, tool calling and thinking mode.

runanywhere/deepseek-v4-1-flash-fastQuantfp8
Input
$0.60/M
Output
$2.40/M
Context
1.0M
Max out
66K
Cache read
$0.01/M
Cache write
$0.60/M
Thinking out
$2.40/M