DeepSeek V4.1 Flash 1017 Ultrafast is now available on Vercel AI Gateway via Fireworks, capable of generating over 1,000 tokens per second. It serves the October 17 snapshot of DeepSeek V4.1 Flash through Fireworks' accelerated serverless infrastructure, with a 1 million token context window. No waitroom, no session time limit, public preview starting now.
The model is built for speed-sensitive use cases: interactive coding, reasoning pipelines, and agentic workflows where latency between generation steps is a bottleneck. It integrates directly with Claude Code, Codex, and Cursor via the Vercel CLI using the model identifier deepseek/deepseek-v4.1-flash-1017-ultrafast. The 1M context window makes it practical for large codebase ingestion, not just short prompts.
The full changelog is worth reading for the AI Gateway infrastructure details: unified API routing, configurable failover, Zero Data Retention support, per-key budget controls, and pass-through provider pricing with no platform markup on inference or BYOK requests. That pricing model is the part most developers will want to verify themselves before committing to production usage.
[READ ORIGINAL →]