Qwen3.6 27B

Qwen3.6-27B is the first open-weight variant in the Qwen3.6 line, released after the Qwen3.5 series and explicitly shaped by community feedback aimed at stability and practical usefulness. The model is positioned as a Causal Language Model with a Vision Encoder, distributed in a Hugging Face Transformers-compatible format that works with vLLM, SGLang, and KTransformers, making it straightforward to drop into existing developer stacks. It carries 27 billion parameters arranged across 64 layers with a 5120 hidden dimension and a 248320-token-padded embedding, and mixes linear and attention blocks in a 16 × 3 × Gated DeltaNet → FFN → 1 × Gated Attention → FFN layout, giving it a hybrid memory design suited for long, multi-turn agent sessions.

The release highlights two practical developer-facing upgrades: more fluent handling of frontend workflows and repository-level reasoning for agentic coding, and a new option to preserve reasoning context across historical messages so iterative tasks do not have to rebuild their chain of thought from scratch. On Blue Claw, the model slots into an OpenAI-compatible endpoint alongside other open models, letting agent workloads such as code review, inbox triage, document indexing, and meeting-note summarization run on lower-cost open weights without changing existing SDK wiring. For teams already comfortable with the Qwen ecosystem, Qwen3.6-27B offers a balanced 27B-scale option that prioritizes responsive coding assistance and sustained multi-step agent behavior over raw flagship scale.