Side-by-side comparison of DeepSeek-V4-Flash (DeepSeek · China) and Llama 4 Maverick (Meta · USA) for self-hosted deployment of the open-weight model. DeepSeek-V4-Flash is rated conditional; Llama 4 Maverick is conditional. They part ways on licence: DeepSeek-V4-Flash is "MIT", Llama 4 Maverick is "Llama community".
| 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 Llama 4 flagship: MoE with 17B active over 128 experts, natively multimodal (text + images). Same Llama community licence as the family: 700M MAU cap, acceptable-use policy, 'Built with Llama' attribution. |
| Last reviewed | 2026-04-28 | 2026-04-15 |
| Open-weight | ||
| Licence | MIT | Llama community |
| Commercial use | Unrestricted | With caps |
| Training data | Categories only | Undisclosed |
| Origin | China | USA |
| Performance & pricing? | ||
| Quality index | 47/100 | 18/100 |
| Speed | 79 tok/s | 116 tok/s |
| Blended price | $0.17/M | $0.50/M |
| Context window | — | — |
| Evidence | ||
| Sources | ||
No overlapping sources between the two entries.