DeepSeek V4 Flash API Hits Public Beta at $0.18/M Tokens
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- DeepSeek rolled out the official V4 Flash API in public beta on July 31, touting enhanced agent capabilities and benchmark scores 'far surpassing' its V4 Pro Preview, per Bloomberg.
- The V4 Flash model, at roughly 300B parameters and the same architecture as the preview, jumped from 61.8 to 82.7 on Terminal-Bench and 7.3 to 54.4 on DeepSwe — all at $0.18 per million output tokens.
- DeepSeek says V4 Pro's official release is 'coming ASAP,' and the Flash upgrade applies only to the API, leaving the Pro API and App/Web models unchanged for now.
- A hacker used DeepSeek AI to autonomously attack vulnerable servers, per BleepingComputer — a security incident that lands the same day the company markets its enhanced agent capabilities.
- Nikkei Asia frames the release as intensifying an AI price war, with Superintelligence reporting DeepSeek 'answered OpenAI's price cut overnight.'
Why it matters: At $0.18 per million output tokens, V4 Flash is being called 'nearly 100x cheaper' than Opus-tier models while matching them on agent benchmarks — pressuring OpenAI, Anthropic, and other API providers to justify their premiums, while BleepingComputer's report of the model being weaponized for autonomous server attacks shows agent capabilities are already being exploited.
