DeepSeek V4 Flash official release: big agent gains, cost advantage
数字生命卡兹克 · wechat · 2026-07-31
WeChat article by Kazike shares experience with DeepSeek V4 Flash official release. Model structure unchanged but post-training greatly improved, with scores rising on TerminalBench2.1, DeepSWE, DSBench-FullStack, and beating V4 Pro Preview on all nine agent tests. Native Responses API support. Author notes gap with top models like GPT-5.6 Sol in hardest coding tasks, but Flash is extremely cost-effective for high-volume routine tasks. His AIHOT consumes 1000 RMB/month in API costs; top models would cost 10x more.
Related event: DeepSeek-V4-Flash GA Launches with Major Agent Capability Boost(17 posts)→
More from coding & agent
- Vercel Launches AI SDK for Python, Snags the 'ai' Namespace — m4rkmc · 2026-07-31
- Model Selection is Becoming Org Design: Structuring AI Workflows — every · 2026-07-31
- Burning a $100 Weekly Limit in One Prompt: The High Cost of Codex Testing — burny_tech · 2026-07-31
- Testing Inkling Small: A Vision-Equipped Model That Can Build Flappy Bird — LiTianleli · 2026-07-31
- Developer Uses Codex Ultra Mode to Bulk-Improve Agent Product, Adds Video Calls — jasonkneen · 2026-07-31
- Open Source Claude Code Skill: Generate Nothing's Minimalist UI Instantly — tom_doerr · 2026-07-31