Brief, mostly signal

DeepSeek V4.1-Flash-1017 adds coding and efficiency gains

Released 17 Oct as a minor update with 1M context, up to 384K output and unchanged API pricing. Weights are on Hugging Face and ModelScope.

Read the source at note.com

What the research found

DeepSeek shipped V4.1-Flash-1017 around Oct 17 as a coding-focused minor update with 1M context, up to 384K output, unchanged API pricing and MIT-licensed weights on Hugging Face and ModelScope; benchmark gains are self-reported by DeepSeek.

Claims and how well the evidence supports them

DeepSeek released V4.1-Flash-1017 around October 17 2026 (date only loosely confirmed) 49%
The model has a 1 million token context and up to 384K output 98%
API pricing is unchanged from before the update 97%
Weights are released on Hugging Face and ModelScope under MIT license 96%

Evidence

note.com https://note.com/kawara_works/n/nf183b1e4833b?hl=en

Actionability rated medium.