DeepSeek-V4-Flash
DeepSeek-V4-Flash is a publicly available API model in public beta with improved agent and coding-agent performance compared to the prior V4-Pro-Preview. The release is a re-post-trained version of the preview that natively supports the Responses API and specific Codex-oriented configurations.
What it does
Developers can call the deepseek-v4-flash model via the API for advanced agent workflows and code-generation/automation tasks.
- Availability
- public beta (official release 2026-07-31)
- Pricing model
- Not established
- Deployment
- hosted API
- Privacy
- Not established
- Licence
- Not established
Capability
Modality
Access
Deployment
Pricing
Target user
Evidence
6 claims- The DeepSeek-V4-Flash API is in public beta and available by setting the model name to deepseek-v4-flash. [1]
The API calling method remains unchanged — simply set the model name to deepseek-v4-flash to use the latest version.
DeepSeek. Quoted text is outside CC BY 4.0 and keeps its original rights. - Agent capabilities are significantly enhanced and outperform V4-Pro-Preview on listed public and internal benchmarks. [1]
Significantly enhanced agent capabilities, with benchmark results far exceeding V4-Pro-Preview:
DeepSeek. Quoted text is outside CC BY 4.0 and keeps its original rights. - Benchmark scores reported include Terminal Bench 2.1: 82.7, NL2Repo: 54.2, Cybergym: 76.7, and DSBench-FullStack: 68.7. [1]
- The model was evaluated on Code Agent tasks using the DeepSeek Harness minimal mode with max effort, topp=0.95, and temperature=1.0. [1]
Note 1: For the Code Agent tasks in the public benchmark sets, the official DeepSeek-V4-Flash was tested using the DeepSeek Harness minimal mode (to be released soon) as the framework, with the max effort level, topp=0.95, and temperature=1.0
DeepSeek. Quoted text is outside CC BY 4.0 and keeps its original rights. - DeepSeek-V4-Flash natively supports the Responses API format and is specifically adapted for Codex configurations. [1]
The official V4-Flash natively supports the Responses API format and is specifically adapted for Codex.
DeepSeek. Quoted text is outside CC BY 4.0 and keeps its original rights. - DeepSeek-V4-Flash-0731 keeps the same model architecture and size as DeepSeek-V4-Flash-Preview but was re-post-trained. [1]
DeepSeek-V4-Flash-0731 keeps the same model architecture and size as DeepSeek-V4-Flash-Preview, and was only re-post-trained.
DeepSeek. Quoted text is outside CC BY 4.0 and keeps its original rights.
Sources
1
1
DeepSeekOwner-controlled
https://api-docs.deepseek.com/updates#2026-07-31
Retrieved 2026-10-28 00:39 UTC
Verified
Versions
1| Version | Effective | Change | Changed fields | State |
|---|---|---|---|---|
| 1 | 2026-07-31 00:00 UTC | publication | None | Current |
In briefs
2- DeepSeek-V4-Flash leads 3 AI developments in week 31, 2026 Weekly brief 2026-W31
- DeepSeek-V4-Flash from DeepSeek, the one AI development on 31 July 2026 Daily brief 2026-07-31
Original catalog content is licensed under CC BY 4.0, attributed to wowsuchai. Source quotations and linked content keep their original rights.