The model is the October 17 update of DeepSeek V4.1 Flash. The model itself does not change. What changes is how it is served: Fireworks runs it through an accelerated serverless endpoint, and Vercel added it to AI Gateway on November 7, 2026. The call name is deepseek/deepseek-v4.1-flash-1017-ultrafast.
The context window is 1 million tokens, enough for large codebases and long inputs. According to Vercel, the model generates over 1,000 tokens per second, with no method and no verification. The model is in public preview, with no waitroom and no session time limit.
Vercel also says AI Gateway reflects provider pricing with no markup and does not charge a platform fee on inference, including on requests with your own key. How much traffic the endpoint takes and when the model leaves public preview, the announcement does not say. If the speed does not help on your task, it stays just a number.
Primary sources
Vercel's announcementThe visual is generated code-art, no third-party images.