Tooony133/Qwen-3.6-27B-HyperBloom

VISIONConcurrent Unit Cost:2Model Size:27BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jul 22, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

Qwen3.6-27B is a 27 billion parameter causal language model developed by Qwen, featuring a vision encoder and a native context length of 262,144 tokens, extensible up to 1,010,000 tokens. This model is specifically optimized for agentic coding tasks, including frontend workflows and repository-level reasoning, and introduces a unique 'Thinking Preservation' feature to streamline iterative development. It demonstrates strong performance across various coding agent benchmarks like SWE-bench and Terminal-Bench 2.0, as well as robust capabilities in knowledge, STEM, reasoning, and multimodal understanding.

Loading preview...

Qwen3.6-27B: Flagship-Level Coding and Multimodal Capabilities

Qwen3.6-27B is a 27 billion parameter causal language model developed by Qwen, building upon the Qwen3.5 series with a focus on stability and real-world utility. It integrates a vision encoder and boasts a native context length of 262,144 tokens, which can be extended up to 1,010,000 tokens using YaRN scaling techniques.

Key Capabilities

  • Enhanced Agentic Coding: Significantly improved handling of frontend workflows and repository-level reasoning, making it highly proficient for complex coding tasks.
  • Thinking Preservation: A novel feature that retains reasoning context from historical messages, reducing overhead and streamlining iterative development, particularly beneficial for agent scenarios.
  • Multimodal Understanding: Supports text, image, and video inputs, demonstrating strong performance across various vision language benchmarks including MMMU, MathVista, and video understanding tasks like VideoMME.
  • Robust Performance: Achieves competitive scores on coding agent benchmarks (e.g., 77.2 on SWE-bench Verified, 59.3 on Terminal-Bench 2.0), knowledge benchmarks (e.g., 86.2 on MMLU-Pro), and STEM & Reasoning tasks (e.g., 93.8 on HMMT Feb 25).
  • Flexible Deployment: Compatible with popular inference frameworks like Hugging Face Transformers, vLLM, SGLang, and KTransformers, with specific optimizations for tool use and Multi-Token Prediction (MTP).

Good For

  • Software Development: Ideal for developers requiring advanced code generation, debugging, and repository-level understanding, especially with its agentic coding and thinking preservation features.
  • Multimodal Applications: Suitable for tasks involving the analysis and generation of content from diverse inputs, including images and videos.
  • Complex Problem Solving: Excels in STEM and reasoning-intensive domains, making it valuable for research and advanced analytical applications.
  • Iterative Development Workflows: The 'Thinking Preservation' feature makes it highly effective for long, iterative development cycles where maintaining context is crucial.