Side-by-side comparison of Codestral 22B (Mistral AI · France) and Qwen3-8B (Alibaba Cloud (Qwen) · China) for self-hosted deployment of the open-weight model. Codestral 22B is rated conditional; Qwen3-8B is conditional. They part ways on licence: Codestral 22B is "MNPL (non-prod)", Qwen3-8B is "Apache 2.0".
| Field | ||
|---|---|---|
| Summary | ||
| Verdict | Conditional EU code model trained on 80+ languages. Licensed under Mistral Non-Production License — blocked for any production or commercial deployment without a paid commercial licence. Use Codestral Mamba (Apache 2.0) if you need commercial freedom. | Conditional Based on published licence terms, Qwen3-8B is released under standard Apache 2.0 with no field-of-use carve-outs, making self-hosted commercial deployment viable. Training-data disclosure is limited to a token count and Chinese origin creates EU AI Act Art. 53 transparency and data-transfer risks that deployers should document. |
| Last reviewed | 2026-04-15 | 2026-04-17 |
| Open-weight | ||
| Licence | MNPL (non-prod) | Apache 2.0 |
| Commercial use | Paid licence req. | Unrestricted |
| Training data | Partial | Token count only |
| Origin | EU | China (Hangzhou) |
| Performance & pricing? | ||
| Quality index | — | 11/100 |
| Speed | — | 86 tok/s |
| Blended price | — | $0.31/M |
| Context window | — | — |
| Evidence | ||
| Sources | ||
No overlapping sources between the two entries.