Qwen3.6-27B
Alibaba (Qwen)
Per the published Apache 2.0 licence, the Qwen3.6-27B weights are deployable without commercial restriction, including for vision-language and 1M-context workloads. The blockers for regulated EU use are the China-based vendor and the absence of any training-data disclosure on the model card — both should be mitigated through self-hosting and a deployer-prepared GPAI compliance file.
Sovereignty
Licence: Apache 2.0Commercial: UnrestrictedTraining data: UndisclosedOrigin: China
Licence facts
- Parameters
- 27B (dense)
- Architecture
- Hybrid Gated DeltaNet + Gated Attention with vision encoder
- Context window
- 262K native, up to ~1M with RoPE/YaRN scaling
- Modalities
- Text, image, video → text
- Reasoning
- Native thinking mode with <think> blocks
Performance & pricing?
Quality index46/100Speed66 tok/sBlended price$1.35/M
Known risks
- Vendor jurisdiction is the People's Republic of China — non-adequate under GDPR Article 45.
- Training corpus is not disclosed on the model card; this complicates AI Act Article 53(1)(d) summary-of-content obligations.
- Vision-language pipeline introduces additional considerations under Article 5 (prohibited practices) and Article 50 (transparency) of the AI Act for any deployment processing biometric or facial imagery.
Sources
Reviewed by Ali Madjaji · Last reviewed 2026-04-28· Reviewed 0 days agoSuggest a correction