DeepSeek V4 Pro 0813 quietly released
DeepSeek has quietly released V4 Pro 0813, an update to its API that adds support for the Responses API format, enabling seamless integration with Codex. The new API supports streaming with server-sent events, function calling, and web search, while maintaining compatibility with existing OpenAI SDK clients. Notably, DeepSeek V4 Pro 0813 introduces a more efficient model, deepseek-v4-flash, and offers automatic context caching, making it a compelling option for developers seeking high-performance AI at scale.
With a simple configuration, you can use DeepSeek models in Codex.
- margalabargala
Right, so quiet it's two of the top three links on HN right now.
- HiPHInch
- storus
Probably nothing Earth-shattering in benchmarks? Flash can already replace Opus 4.5 for local agentic coding, maybe they didn't have a similar win for Pro?