Summary
On August 12, 2026, DeepSeek moved its flagship V4-Pro (V4-Pro-0813) to general availability across web, mobile, and API. The 1.6-trillion-parameter mixture-of-experts model (about 49B active per token) supports a 1M-token context and up to 384K output tokens, tuned for agentic tool use and multi-step workflows. A V4 family price increase with peak/off-peak billing took effect August 16.
What changed
DeepSeek released V4-Pro-0813 to GA on web, app, and API: 1.6T total / ~49B active parameters, a 1,048,576-token context, and up to 384K output tokens, optimized for tool use and multi-step agent tasks. On Aug 16, DeepSeek introduced peak/off-peak V4 pricing, raising V4-Pro output to $3.96 per million at peak (off-peak at half).
Why it matters
A frontier-scale, agent-tuned open model with a 1M context and large output budget strengthens the low-cost open alternative to US frontier labs; the new peak/off-peak pricing signals capacity pressure as agent workloads scale.
Evidence excerpt
"DeepSeek V4 Pro 0813 reached general availability... The model has 1.6 trillion total parameters with roughly 49 billion active per token and supports a 1,048,576-token context window and up to 384,000 output tokens." (benchmark gains vendor-reported)