DeepSeek Launches V4-Flash API: Massive Agent Gains at $0.14/M Input Tokens
dair_ai · x · 2026-07-31
DeepSeek has officially launched the public beta for its V4-Flash model API. The new version features massively upgraded Agent capabilities, with benchmark scores far surpassing the previous V4-Pro-Preview—jumping over 20 points on TerminalBench-2.1.
Additionally, the official V4-Flash now natively supports the Responses API format and is fully adapted for Codex. Its pricing is aggressively low: input costs are only $0.14 per million tokens, and output is $0.28. This signals a fierce battle to make AI intelligence "too cheap to meter."
Related event: DeepSeek V4-Flash Official Release Boosts Agent Capabilities(75 posts)→
More from Models
- Users Report Claude's Writing Has Become Weird and Hard to Read — altryne · 2026-08-01
- DeepSeek-V4-Flash Enters Public Beta, Outperforming V4-Pro-Preview in Benchmarks — petrusenko_max · 2026-08-01
- DeepSeek's Chain of Thought Exclaims "OH MY GOD" in Viral Trace — fragment_me · 2026-08-01
- Developer Complains Claude's English Writing Style Has Severely Degraded — smolix · 2026-08-01
- Hugging Face Attacked by Secret Proprietary Models, Defended by Open Source — _akhaliq · 2026-08-01
- Hands-on: DeepSeek V4 Flash Stays Coherent at 200K Context, Excels in Reasoning — Nyghtbynger · 2026-08-01