Side-by-side comparison of Gemma 4 E4B Instruct (Google DeepMind · United States) and Kimi-K2.6 (Moonshot AI · China) for self-hosted deployment of the open-weight model. Gemma 4 E4B Instruct is rated conditional; Kimi-K2.6 is conditional. They part ways on licence: Gemma 4 E4B Instruct is "Apache 2.0", Kimi-K2.6 is "Modified MIT".
| Field | ||
|---|---|---|
| Summary | ||
| Verdict | Conditional Based on published licence terms, Gemma 4 E4B is an edge-optimised multimodal variant under pure Apache 2.0 with no prohibited-use carve-outs. Audio input (30s) and on-device deployment push GDPR biometric, AI Act emotion-recognition, and Art. 25 data-protection-by-design obligations entirely onto the integrator with no Google-side telemetry or kill-switch. | Conditional Per the published LICENSE file, Kimi-K2.6 ships under a Modified MIT licence: identical to standard MIT for almost all deployers, with an additional UI-attribution requirement only for products exceeding 100M monthly active users or USD 20M monthly revenue. The genuine EU-readiness concerns are the China-based vendor and the absence of any training-data disclosure rather than the licence itself. |
| Last reviewed | 2026-04-17 | 2026-04-28 |
| Open-weight | ||
| Licence | Apache 2.0 | Modified MIT |
| Commercial use | Unrestricted | Attribution at scale |
| Training data | Domain-level summary | Undisclosed |
| Origin | United States | China |
| Performance & pricing? | ||
| Quality index | 15/100 | 54/100 |
| Speed | — | — |
| Blended price | — | $1.71/M |
| Context window | — | — |
| Evidence | ||
| Sources | ||
No overlapping sources between the two entries.