CrowdMind/Qwen3.5-9b-pro

VISIONPricing:Input $0.431 / Cached $0.0862 / Output $1.12Concurrent Unit Cost:1Model Size:9BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Sep 22, 2026License:mitArchitecture:Transformer Open Weights Featherless Exclusive Cold

CrowdMind/Qwen3.5-9b-pro is a 9 billion parameter supervised fine-tuned version of Qwen/Qwen3.5-9B, developed by CrowdMind and Dustin Loring. This model is designed for agentic workflows, coding, tool use, automation, and general-purpose reasoning. It retains the multimodal message structure of the Qwen3.5 family, supporting image and video inputs with a 32768 token context length. Benchmarks show significant improvements over its base model in code, automation, and tool-use tasks.

Loading preview...

CrowdMind Qwen3.5-9B-Pro Overview

CrowdMind/Qwen3.5-9b-pro is a 9 billion parameter model, a supervised fine-tuned (SFT) version of Qwen/Qwen3.5-9B, developed by CrowdMind and Dustin Loring. It is specifically engineered for advanced agentic workflows, code generation, tool utilization, and automation tasks, while also excelling in general reasoning.

Key Capabilities

  • Enhanced Agentic Performance: Demonstrates substantial improvements in agentic task execution, with significant gains on benchmarks like SWE Pro (from 32.0 to 44.6 avg@3) and AutomationBench (from 5.0 to 30.3 avg@1).
  • Multimodal Understanding: Supports both image and video inputs, leveraging the Qwen3.5 family's multimodal message structure. It uses a custom chat template for handling structured multimodal content.
  • Configurable Reasoning: Features explicit controls for reasoning effort (low, medium, xhigh) via its custom chat template, allowing users to tailor the model's thinking process.
  • Tool Use & Automation: Optimized for integrating with external tools and automating complex tasks, as evidenced by improved scores on Toolathlon-Verified and Terminal Bench.
  • Structured Responses: Designed to produce structured outputs, crucial for agentic and tool-use applications.

Good For

  • Developers building agentic systems requiring robust task execution and planning.
  • Code generation and modification in various programming contexts.
  • Applications needing multimodal understanding of images and videos.
  • Automation of complex workflows and terminal-oriented tasks.
  • Scenarios demanding controlled reasoning and structured outputs.

Finetunes
1 models