AgentFlow/agentflow-planner-3b

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:3.1BQuant:BF16Ctx Length:32kPublished:Oct 6, 2025License:mitArchitecture:Transformer0.0K Open Weights Warm

AgentFlow/agentflow-planner-3b is a 3.1 billion parameter instruction-tuned model developed by AgentFlow, built upon the Qwen2.5-3B-Instruct architecture. This model is specifically designed as a planner agent, excelling at generating structured plans for complex tasks. With a notable 32768 token context length, it is optimized for reasoning and sequential decision-making in agentic workflows.

Loading preview...

AgentFlow Planner Agent 3B Overview

AgentFlow/agentflow-planner-3b is a 3.1 billion parameter model, based on the Qwen2.5-3B-Instruct architecture, specifically fine-tuned to function as a planner agent. Developed by AgentFlow, this model is designed to generate structured and coherent plans for various tasks, making it suitable for integration into larger agentic systems.

Key Capabilities

  • Task Planning: Specializes in breaking down complex goals into actionable steps.
  • Instruction Following: Built upon an instruction-tuned base, ensuring robust adherence to prompts.
  • Extended Context: Features a 32768 token context window, allowing for detailed problem descriptions and multi-step planning.

Good For

  • Agentic Workflows: Ideal for use as the planning component within multi-agent architectures.
  • Complex Task Automation: Generating sequential plans for automated systems.
  • Reasoning Tasks: Leveraging its planning capabilities for structured problem-solving.

For more technical details, refer to the HF Paper and the AgentFlow GitHub repository.