88Z AI, 31 July — DeepSeek has launched the public beta for the official DeepSeek-V4-Flash API. This release features significantly enhanced agent capabilities, outperforms V4-Pro-Preview in benchmarks, natively supports the Responses API format, and is optimized for Codex, while maintaining the original model structure with only post-training updates.
At a glance
- Agent capabilities are significantly enhanced, with benchmark scores far exceeding V4-Pro-Preview
- Natively supports Responses API and is optimized for Codex, while keeping the model structure unchanged
The update
DeepSeek has officially announced the public beta launch of the DeepSeek-V4-Flash official API through an update to its API documentation release logs. The core highlight of this update is a significant enhancement in agent capabilities, with scores in multiple benchmark tests substantially surpassing the previous V4-Pro-Preview version. In terms of technical compatibility, the official release natively supports the Responses API format and has been specifically optimized and adapted for Codex to provide a smoother development experience. Notably, the underlying model structure of DeepSeek-V4-Flash remains identical to the Preview version, with no architectural changes made. The primary improvements stem from a renewed post-training process, achieving a performance leap while maintaining structural stability.
Sources & editorial note
This article is an AI-assisted synthesis by 88Z AI based on the public sources below. Product claims and availability are attributed to their publishers. It does not represent independent testing or first-hand reporting.
- IT之家Source published
Source dates above refer to the original announcements; the date under the headline is this article’s publication on 88Z AI.