MuiRouterDeepSeek V4 Pro Is Now Generally Available, Now on MuiRouter
DeepSeek shipped the V4 Pro stable (V4-Pro-0813): 1.6T MoE, 1M context, 384K max output, and thinking/non-thinking modes. Now live on MuiRouter with one API key.
DeepSeek has shipped the stable version of its flagship — DeepSeek-V4-Pro-0813 (deepseek-v4-pro) — completing the V4 family that began with the April preview. The official production release keeps the 1.6T total / 49B active MoE architecture and the family's signature 1M-token context window, with a maximum output of 384K tokens.
deepseek-v4-pro is now live on MuiRouter. With a single API key and a simple model name change, you can call it through the standard OpenAI-compatible endpoint (/v1/chat/completions).
Key Highlights
| Feature | DeepSeek V4 Pro |
|---|---|
| Architecture | 1.6T total parameters MoE, 49B active per token |
| Context Window | 1,048,576 tokens (1M) |
| Max Output | up to 384K tokens |
| Thinking Mode | non-thinking / thinking (default) |
| Capabilities | JSON Output, Tool Calls, Responses API, Anthropic API, Chat Prefix, FIM (non-thinking only) |
Official Pricing & MuiRouter Billing
DeepSeek official API rates:
| Token Type | Price (per 1M tokens) |
|---|---|
| Input (Cache Miss) | $0.435 |
| Input (Cache Hit) | $0.003625 |
| Output | $0.87 |
On MuiRouter, deepseek-v4-pro carries our standard 1.2 markup: $0.522 / $0.00435 / $1.044 per 1M tokens (input, cached input, output). Note that DeepSeek has officially announced an overall price increase in the near future — a good reason to lock in current rates.
Conclusion
The open-weight flagship of the DeepSeek V4 family is now generally available: 1M context, thinking and non-thinking modes, and aggressive prompt-caching economics at a fraction of the cost of comparable closed models. Try it today in the MuiRouter Playground, in your terminal AI agent, or through the OpenAI-compatible API — transparent pay-as-you-go billing, no monthly subscription.
Reference sources
Primary source published on August 12, 2026.
Be ready for the next AI shift
Start with one API key and a cleaner path to keep model access stable as tools and upstream availability change.