DeepSeek-V4-Flash Official API Released with Major Agent Upgrades
APPSO · wechat · 2026-07-31
DeepSeek has launched the public beta of the V4-Flash official API, with the core highlight being a massive leap in Agent capabilities.
The new model significantly outperforms the previous V4-Pro-Preview across multiple benchmarks: scoring 82.7 on TerminalBench2.1, 76.7 on Cybergym, and 70.3 on ToolathlonVerified. It also achieved 68.7 and 59.6 on the internal full-stack development test DSBench-FullStack and the high-difficulty coding test DSBench-Hard, respectively.
Additionally, the official release natively supports the Responses API and is fully adapted for Codex. Once configured, it can be seamlessly used across Codex CLI, the ChatGPT desktop app, and the VSCode Codex plugin. Currently, only V4-Flash supports Codex integration, with V4-Pro expected to follow in early August.
Notably, this capability boost is primarily attributed to renewed post-training. The update is currently limited to the V4-Flash API, with the App, Web, and V4-Pro API remaining unchanged. The official release of the more powerful V4-Pro will arrive at a later date.
Related event: DeepSeek-V4-Flash GA Launches with Major Agent Capability Boost(17 posts)→
More from coding & agent
- Vercel Launches AI SDK for Python, Snags the 'ai' Namespace — m4rkmc · 2026-07-31
- Model Selection is Becoming Org Design: Structuring AI Workflows — every · 2026-07-31
- Burning a $100 Weekly Limit in One Prompt: The High Cost of Codex Testing — burny_tech · 2026-07-31
- Testing Inkling Small: A Vision-Equipped Model That Can Build Flappy Bird — LiTianleli · 2026-07-31
- Developer Uses Codex Ultra Mode to Bulk-Improve Agent Product, Adds Video Calls — jasonkneen · 2026-07-31
- Open Source Claude Code Skill: Generate Nothing's Minimalist UI Instantly — tom_doerr · 2026-07-31