Side-by-side comparison of DeepSeek-V4-Flash (DeepSeek · China) and Kimi K2 Instruct (Moonshot AI · China) for self-hosted deployment of the open-weight model. DeepSeek-V4-Flash is rated conditional; Kimi K2 Instruct is conditional. They part ways on licence: DeepSeek-V4-Flash is "MIT", Kimi K2 Instruct is "Modified MIT".
| Field | ||
|---|---|---|
| Summary | ||
| Verdict | 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. | Conditional 1T-parameter MoE (32B active) tuned for agentic and tool-use workflows. Modified MIT permits commercial use. Same Chinese-origin alignment and supply-chain considerations as DeepSeek and Qwen. |
| Last reviewed | 2026-04-28 | 2026-04-16 |
| Open-weight | ||
| Licence | MIT | Modified MIT |
| Commercial use | Unrestricted | Yes |
| Training data | Categories only | Undisclosed |
| Origin | China | China |
| Performance & pricing? | ||
| Quality index | 47/100 | 26/100 |
| Speed | 79 tok/s | 34 tok/s |
| Blended price | $0.17/M | $1.04/M |
| Context window | — | — |
| Evidence | ||
| Sources | ||
No overlapping sources between the two entries.