Side-by-side comparison of Gemma 4 26B A4B Instruct (Google DeepMind · United States) and MiMo-V2.5-Pro (Xiaomi (MiMo)) for self-hosted deployment of the open-weight model. Gemma 4 26B A4B Instruct is rated conditional; MiMo-V2.5-Pro is conditional. They part ways on licence: Gemma 4 26B A4B Instruct is "Apache 2.0", MiMo-V2.5-Pro is "MIT".
| Field | ||
|---|---|---|
| Summary | ||
| Verdict | Conditional Based on published licence terms, Gemma 4 26B A4B ships under pure Apache 2.0 with no prohibited-use carve-outs — a departure from prior Gemma generations. The sparse-MoE architecture (25.2B total / 3.8B active) puts it in an ambiguous zone for EU AI Act GPAI systemic-risk classification, and US origin plus image-input support add transparency obligations that deployers should document. | Conditional Per the published LICENSE file, MiMo-V2.5-Pro ships under MIT, so the weights themselves carry no commercial restriction. The remaining EU-readiness gaps are the China-based vendor and the corpus disclosure that names training-stage categories (text pre-training, multimodal pre-training, SFT, RL, MOPD) without listing datasets — both should be addressed in any GPAI deployer file before regulated use. |
| Last reviewed | 2026-04-17 | 2026-04-28 |
| Open-weight | ||
| Licence | Apache 2.0 | MIT |
| Commercial use | Unrestricted | Unrestricted |
| Training data | Domain-level summary | Categories only |
| Origin | United States | China |
| Performance & pricing? | ||
| Quality index | 27/100 | 54/100 |
| Speed | — | 65 tok/s |
| Blended price | — | $1.50/M |
| Context window | — | — |
| Evidence | ||
| Sources | ||
No overlapping sources between the two entries.