DeepSeek-V4-Flash Update

	DeepSeek-V4-Flash Update(api-docs.deepseek.com)
	737 points by dnhkng 45 days ago \| 345 comments
	tl;dr: DeepSeek-V4-Flash is now in public beta, accessible via the `deepseek-v4-flash` model name, with the same architecture as the Preview but re-post-trained for significantly stronger agent capabilities (e.g., Terminal Bench 2.1: 82.7, SWE-related benchmarks well above V4-Pro-Preview). It natively supports the Responses API format and is specifically adapted for Codex. The V4-Pro API and app/web models are unchanged, with an official V4-Pro release coming soon.
	HN Discussion: ↑Flash model's efficiency and low cost make capability improvements highly impactful for real-world tasks ↑Personal usage confirms Flash is excellent for coding/agent tasks, often preferred over Pro or frontier models ↑The benchmarks are impressive, with a smaller model outperforming much larger competitors •Trust in DeepSeek regarding data/training consent sets it apart from other providers ↓Questioning the versioning/naming choice rather than calling it v4.1-Flash