Side-by-side comparison of Gemma 4 26B A4B Instruct (Google DeepMind · United States) and MiniCPM5-1B (OpenBMB) for self-hosted deployment of the open-weight model. Gemma 4 26B A4B Instruct is rated conditional; MiniCPM5-1B is conditional. They part ways on training data: Gemma 4 26B A4B Instruct is "Domain-level summary", MiniCPM5-1B is "Documented".
| 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 Apache 2.0 LICENSE, the 1.08B-parameter weights carry no use restrictions and are well suited to on-device or edge deployments. Unusually for a Chinese-origin release, OpenBMB ships a full data card naming every pre-training and SFT corpus, which closes the AI Act Art. 53 transparency gap most peers leave open. The remaining hedge for regulated EU buyers is vendor jurisdiction (Tsinghua-affiliated lab, mainland China) — self-host the weights and treat any vendor-side service as out of scope. |
| Last reviewed | 2026-04-17 | 2026-05-30 |
| Open-weight | ||
| Licence | Apache 2.0 | Apache 2.0 |
| Commercial use | Unrestricted | Unrestricted |
| Training data | Domain-level summary | Documented |
| Origin | United States | China |
| Performance & pricing? | ||
| Quality index | 27/100 | — |
| Speed | — | — |
| Blended price | — | — |
| Context window | — | — |
| Evidence | ||
| Sources | ||
No overlapping sources between the two entries.