Atomic-Germ/Qwen3.6-35B-A3B-NPU2

TEXT GENERATIONConcurrent Unit Cost:3Model Size:35.1BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jul 11, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

Qwen3.6-35B-A3B is a 35 billion parameter causal language model with a vision encoder developed by Qwen. This model, with 3 billion activated parameters, is optimized for agentic coding, handling frontend workflows and repository-level reasoning with enhanced fluency and precision. It features a native context length of 262,144 tokens, extensible up to 1,010,000 tokens, and introduces thinking preservation to streamline iterative development.

Loading preview...

Model Overview

Qwen3.6-35B-A3B is a 35 billion parameter causal language model with a vision encoder, developed by Qwen. It builds upon the Qwen3.5 series, focusing on stability and real-world utility for developers. The model has 3 billion activated parameters and supports a native context length of 262,144 tokens, which can be extended up to 1,010,000 tokens using RoPE scaling techniques like YaRN.

Key Capabilities

  • Agentic Coding: Significantly improved handling of frontend workflows and repository-level reasoning, offering greater fluency and precision.
  • Thinking Preservation: A novel feature that retains reasoning context from historical messages, reducing overhead and streamlining iterative development.
  • Multimodal Input: Supports image and video inputs, enabling a wide range of vision-language tasks.
  • High Performance: Achieves competitive results across various benchmarks, including coding agent tasks (e.g., SWE-bench Verified 73.4, Terminal-Bench 2.0 51.5), general agent tasks (e.g., MCPMark 37.0), and vision-language tasks (e.g., MMBench 92.8, OmniDocBench1.5 89.9).
  • Multi-Token Prediction (MTP): Trained with multi-step prediction for enhanced efficiency.

When to Use This Model

  • Coding Agents: Ideal for applications requiring advanced code generation, debugging, and repository-level understanding, especially for frontend development.
  • Iterative Development: Beneficial for scenarios where preserving reasoning context across multiple interactions is crucial for efficiency and consistency.
  • Multimodal Applications: Suitable for tasks involving both text and visual (image/video) inputs, such as visual question answering, document understanding, and video analysis.
  • Long Context Processing: Recommended for applications that require processing and generating very long texts, with support for up to 1,010,000 tokens.