DeepSeek's budget model now speaks OpenAI's language — and beats its own Pro at agent tasks
DeepSeek moved its V4-Flash API from preview to an official public-beta release on July 31. The retrained model keeps the same architecture but now surpasses the company's higher-tier V4-Pro-Preview on several coding-agent benchmarks, scoring 82.7 on Terminal Bench 2.1. It also natively supports OpenAI's Responses API format and is tuned for Codex-style workflows; DeepSeek's consumer app and V4-Pro API are unchanged.
Why it mattersBy speaking OpenAI's API dialect natively, DeepSeek positions its cheapest model as a drop-in replacement inside existing agent pipelines — an answer to this week's OpenAI price cuts and another sign of how fast Chinese labs are commoditizing agent-grade coding models.
▶ Related video: DeepSeek V4 Flash Fully Local — 32 tok/s on a Single Chip
Read in the app — free, in 9 languages
Related stories
Rumour: the anonymous model that just topped a coding benchmark, for free, is said to be Zhipu's unreleased flagship2026-08-21DeepSeek's cheap workhorse can now see — and on agent tasks that need eyes it says it is close to Anthropic's best
2026-08-21Nvidia is paying $6 billion for the machine that builds a rival's models — and hiring 109 of the people who ran it
2026-08-21Given four hours and a GPU to improve the way AI is trained, the best agent scored 0.25 out of 1 — and most never tried
2026-08-21ChatGPT can now read your iMessages and send them — and the setting that lets it skip asking is the one OpenAI warns about
2026-08-21