Side-by-side comparison of Gemma 4 31B Instruct (Google DeepMind · United States) and Qwen 3.5 (Alibaba · China) for self-hosted deployment of the open-weight model. Gemma 4 31B Instruct is rated conditional; Qwen 3.5 is conditional. They part ways on commercial use: Gemma 4 31B Instruct is "Unrestricted", Qwen 3.5 is "Yes".
| Field | ||
|---|---|---|
| Summary | ||
| Verdict | Conditional Based on published licence terms, Gemma 4 31B ships under pure Apache 2.0 — a notable break from the Gemma Terms of Use used in prior generations — with no prohibited-use carve-outs. US origin carries Schrems-II and CLOUD-Act exposure, and at 30.7B dense the model likely crosses EU AI Act GPAI systemic-risk thresholds that enterprise deployers should document. | Conditional Hybrid Gated-DeltaNet + MoE flagship (397B total, 17B active) under Apache 2.0. Native vision, 201 languages, 262K context (1M with YaRN). Licence is clean; Chinese-origin alignment and supply-chain considerations persist. |
| Last reviewed | 2026-04-17 | 2026-04-15 |
| Open-weight | ||
| Licence | Apache 2.0 | Apache 2.0 |
| Commercial use | Unrestricted | Yes |
| Training data | Domain-level summary | Undisclosed |
| Origin | United States | China |
| Performance & pricing? | ||
| Quality index | 32/100 | 40/100 |
| Speed | — | 53 tok/s |
| Blended price | — | $1.35/M |
| Context window | — | — |
| Evidence | ||
| Sources | ||
No overlapping sources between the two entries.