DeepSeek V4.1 Flash: 890 bytes/token KV cache, 552B MoE beats V4 Pro on agentic tasks

togethercompute · x · 2026-09-13

Together AI has launched DeepSeek V4.1 Flash, with full specs now published:

Related event: DeepSeek V4.1 Flash Launches on Together AI, Beating GPT-5.6 Sol at a Third of the Cost(3 posts)→

Original post →

More from coding & agent

coding & agent channel →