DeepSeek-V4-Flash Formal Version Released with Enhanced Agent Capabilities
Johell1NS · ghdev · 2026-07-31
DeepSeek has announced the formal release of DeepSeek-V4-Flash (checkpoint 0731). Retaining the same structure as the preview, this version has been retrained to offer significantly improved agent capabilities.
Key benchmark scores include:
- Terminal Bench: 82.7
- NL2Repo: 54.2
- Cybergym: 76.7
- DeepSWE: 54.4
- Toolathlon: 70.3
It also natively supports the Responses API. Because the model ID remained unchanged, the developer community is currently verifying whether gateway platforms like OpenCode Go/Zen are serving the new checkpoint.
More from Models
- Why are sub-15B open models so rare now? Labs seem to have settled on 24-27B minimum — Hot_Example_4456 · 2026-10-16
- User spots model openly generating "ideal responses" to train other models — liminal_bardo · 2026-10-16
- Reflection's 501B MoE Beam hits 463 tok/s on 8x GB200 with NVFP4 at day 0 — zhyncs42 · 2026-10-16
- Gemma 4 31B QAT on one 24GB card: 46 tok/s plain, ~80 with draft model on RTX 4090 — Subject_Poet_3861 · 2026-10-16
- JevBench v1.7.46 leaderboard: new d1-ultra tops 32 API systems at 76.1 composite — airesearch12 · 2026-10-16
- Team of 5 mulls dropping ChatGPT Pro for Business as OpenAI quota parity stays unclear — bettercallfool · 2026-10-16