Latest / Frontier models and safety
DeepSeek releases V4.1-Flash and routes its V4-Pro traffic to it
See how this story connectsDeepSeek released DeepSeek-V4.1-Flash, a 552-billion-parameter mixture-of-experts model with a new causal encoder-decoder design that uses 8 billion active parameters for input and 16 billion for output, with open weights on Hugging Face. DeepSeek said it beats its flagship V4-Pro on benchmarks, needs a quarter of the HBM for its KV cache, and from 14 September routed all V4-Pro API requests to V4.1-Flash at Flash prices until V4.1-Pro launches.
Why it matters
Developers on DeepSeek’s API moved to a cheaper open-weights model by default, and anyone can download it to run themselves.
- When
- 10 Sep 2026
- ID
WTA-N-2026-0126- Where
- China
- Who
- DeepSeek
Summary by wheresthe.ai, from DeepSeek API docs. Checked against the source on 9 Oct 2026.
Source copy: download the page as saved on 18 Oct 2026 (PDF). Also held by the Internet Archive.
Cite this story
WTA-N-2026-0126: “DeepSeek releases V4.1-Flash and routes its V4-Pro traffic to it.” wheresthe.ai, 10 Sep 2026. https://wheresthe.ai/d/cn-deepseek-v4-1-flash-2026-09/ (source: DeepSeek API docs: V4.1-Flash).
Spotted an error? Email [email protected].