Side-by-side comparison of Gemma 4 26B A4B Instruct (Google DeepMind · United States) and GPT-OSS 20B (OpenAI · USA) for self-hosted deployment of the open-weight model. Gemma 4 26B A4B Instruct is rated conditional; GPT-OSS 20B is conditional. On the four sovereignty dimensions we track, they score identically — the difference is in context and org posture.
| 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 Based on published licence terms, GPT-OSS 20B is released under Apache 2.0 with no field-of-use carve-outs in the licence itself; OpenAI publishes a separate non-binding 'gpt-oss usage policy' as guidance. Training-data disclosure is domain-level only, and US origin carries Schrems-II / CLOUD-Act exposure that enterprise deployers should document for EU workloads. |
| Last reviewed | 2026-04-17 | 2026-04-17 |
| Open-weight | ||
| Licence | Apache 2.0 | Apache 2.0 |
| Commercial use | Unrestricted | Unrestricted |
| Training data | Domain-level summary | Domain-level summary |
| Origin | United States | United States |
| Performance & pricing? | ||
| Quality index | 27/100 | 25/100 |
| Speed | — | 294 tok/s |
| Blended price | — | $0.10/M |
| Context window | — | — |
| Evidence | ||
| Sources | ||
No overlapping sources between the two entries.