DeepSeek has released the general availability build of V4 Pro, version 0813, and it is now live on OpenRouter. The large-scale mixture-of-experts model is drawing intense attention, with the Hacker News thread around the launch passing 640 points and 230 comments within hours of appearing.
The headline specs are a 1,048,576 token context window, up to 384,000 output tokens, and pricing of 43.5 cents per million input tokens and 87 cents per million output tokens. Cache reads cost just 0.36 cents per million tokens, and the OpenRouter listing includes independent benchmarks from Artificial Analysis for buyers who want third-party numbers.
The launch shows DeepSeek moving fast in the frontier-model race, shipping the GA of its V4 generation as rivals push out their own flagship updates. For founders building agentic or long-context workloads, the combination of a million-token window, competitive benchmark scores and a budget price point makes V4 Pro an increasingly attractive default to test against.