Side-by-side comparison of DeepSeek V3.2 (DeepSeek · China) and DeepSeek-V4-Flash (DeepSeek · China) for self-hosted deployment of the open-weight model. DeepSeek V3.2 is rated conditional; DeepSeek-V4-Flash is conditional. They part ways on commercial use: DeepSeek V3.2 is "Yes", DeepSeek-V4-Flash is "Unrestricted".
| Field | ||
|---|---|---|
| Summary | ||
| Verdict | Conditional 685B successor to V3 with DeepSeek Sparse Attention for long context, scalable RL for agentic tasks. Vendor claims parity with GPT-5 (Speciale variant exceeds). MIT licence keeps weights clean; Chinese-origin considerations unchanged. | Conditional DeepSeek-V4-Flash is the smaller-active sibling of V4-Pro under the same MIT terms — permissive on the weights, but the China-based vendor and undocumented training corpus mean any EU deployment still needs a self-hosted topology and a deployer-side GPAI documentation file under AI Act Article 53. |
| Last reviewed | 2026-04-15 | 2026-04-28 |
| Open-weight | ||
| Licence | MIT | MIT |
| Commercial use | Yes | Unrestricted |
| Training data | Undisclosed | Categories only |
| Origin | China | China |
| Performance & pricing? | ||
| Quality index | 32/100 | 47/100 |
| Speed | 32 tok/s | 79 tok/s |
| Blended price | $0.32/M | $0.17/M |
| Context window | — | — |
| Evidence | ||
| Sources | ||
No overlapping sources between the two entries.