Side-by-side comparison of DeepSeek V3.2 (DeepSeek · China) and MiniMax-M2.7 (MiniMax AI · China) for self-hosted deployment of the open-weight model. DeepSeek V3.2 is rated conditional; MiniMax-M2.7 is blocked. They part ways on licence: DeepSeek V3.2 is "MIT", MiniMax-M2.7 is "MiniMax Non-Commercial License".
| 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. | Blocked Per current documentation, the MiniMax Non-Commercial License prohibits commercial deployment without individually negotiated written authorization from MiniMax, making the weights unsuitable for EU commercial workloads out-of-the-box. Opaque training data and Shanghai-based publisher compound the EU AI Act and data-transfer gaps. |
| Last reviewed | 2026-04-15 | 2026-04-17 |
| Open-weight | ||
| Licence | MIT | MiniMax Non-Commercial License |
| Commercial use | Yes | Non-commercial only |
| Training data | Undisclosed | Undisclosed |
| Origin | China | China (Shanghai) |
| Performance & pricing? | ||
| Quality index | 32/100 | 50/100 |
| Speed | 32 tok/s | 46 tok/s |
| Blended price | $0.32/M | $0.53/M |
| Context window | — | — |
| Evidence | ||
| Sources | ||
No overlapping sources between the two entries.