DeepSeek-V4-Flash Officially Releases Public Beta Version, Surpasses Zhipu GLM-5.2 in Performance
17 hour ago / Read about 0 minute
Author:小编   

On July 31, the public beta testing commenced for the official version API of DeepSeek-V4-Flash. As per the official announcement, the latest iteration brings about a substantial improvement in Agent capabilities, with benchmark test results significantly outperforming those of its predecessors, including V4-Pro-Preview, as well as Zhipu GLM-5.2. In particular, it scored 82.7 on Terminal Bench 2.1, 76.7 on Cybergym, and 68.7 and 59.6 on DSBench-FullStack and DSBench-Hard, respectively. Moreover, the official version of V4-Flash now natively supports the Responses API format and maintains compatibility with Codex, thereby streamlining the integration process for developers and cutting down on associated costs. It's important to note that this update is exclusive to the V4-Flash API interface, with no modifications being made to the V4-Pro API or the APP/WEB-end models at this juncture.