DeepSeek-V4-Flash Official API Enters Public Beta, V4-Pro to 'Release Soon'
DeepSeek has announced that the official API of DeepSeek-V4-Flash has entered public beta testing, with significantly enhanced Agent capabilities and benchmark scores far exceeding V4-Pro-Preview.
According to IT Home, on the afternoon of July 31, DeepSeek announced via its API documentation release log that the official API of DeepSeek-V4-Flash has entered public beta testing.
According to the introduction, DeepSeek-V4-Flash's Agent capabilities have been significantly enhanced, with benchmark tests far exceeding V4-Pro-Preview:
Terminal Bench 2.1: 82.7
NL2Repo: 54.2
Cybergym: 76.7
DeepSWE: 54.4
Toolathlon verified: 70.3
Agent Last Exam: 25.2
Automation Bench (Public): 25.1
DSBench-FullStack: 68.7
DSBench-Hard: 59.6
It is reported that the official V4-Flash natively supports the Responses API format and is specifically adapted for Codex. The model structure and size of DeepSeek-V4-Flash-0731 remain the same as DeepSeek-V4-Flash-preview, with only post-training redone. IT Home provides the link to the specific configuration documentation: Click here
Officials also stated that this time only the DeepSeek-V4-Flash API interface was upgraded. The DeepSeek-V4-Pro API and the APP/WEB models have not been changed, and the official version of DeepSeek-V4-Pro will be released as soon as possible.

评论
0 条讨论
登录后发表评论
立即登录暂无评论
成为第一个分享想法的人吧!