27B GGUF Released

Unsloth released Qwen3.6-27B-GGUF along with a running and fine-tuning guide.

Qwen3.6 is a 27B open-weight model with enhanced agentic coding and thinking preservation, using a Causal LM architecture that includes a vision encoder.

• Context length: 262,144 tokens by default, expandable up to 1,010,000 tokens
• Runtime support: Examples provided for SGLang, vLLM, KTransformers, and Hugging Face Transformers
• API usage: Includes OpenAI-compatible Chat Completions call examples
• Improved tool calling: Designed to improve tool calling success rate through better nested object parsing
• Unsloth Studio: Lets you run and fine-tune Qwen3.6, and also check GGUF quantization benchmarks

For production deployment, use the latest framework versions, and dedicated serving engines such as SGLang, KTransformers, and vLLM are recommended for production environments.