seanpoyner/smolcode-coder-powershell-3b-tools

TEXT GENERATIONConcurrency Cost:1Model Size:3.1BQuant:BF16Ctx Length:32kTool Calling:SupportedPublished:Jun 14, 2026License:apache-2.0Architecture:Transformer Open Weights Cold

The seanpoyner/smolcode-coder-powershell-3b-tools model is a 3.1 billion parameter LoRA fine-tune of Qwen2.5-Coder-1.5B-Instruct, specifically designed to enable native function call emission. This model addresses the limitation of smaller coder models that typically output tool calls as plain-text JSON, thereby facilitating agentic coding loops. It is optimized for driving agentic coding assistants by correctly parsing tool-use trajectories.

Loading preview...

Overview

This model, seanpoyner/smolcode-coder-powershell-3b-tools, is a LoRA fine-tune of the Qwen2.5-Coder-1.5B-Instruct base model, specifically engineered to enable native <tool_call> function call emission. It was developed for the smolcode agentic coding assistant during the Hugging Face "Build Small" hackathon. The primary goal is to allow tiny (\u22642B parameter) coder models to effectively drive agentic coding loops by correctly formatting tool calls, which standard small Qwen-Coder models typically output as unparseable plain-text JSON.

Key Capabilities

  • Native Tool Call Emission: Fine-tuned to output tool calls in the <tool_call>{"name": ..., "arguments": ...}</tool_call> format, compatible with runtimes like Ollama and llama.cpp.
  • Agentic Loop Integration: Closes the gap for small language models (SLMs) to participate in agentic tool-use workflows.
  • Optimized Training: Utilizes bf16 LoRA with assistant-only loss on tool calls and final answers, trained on a combination of NousResearch/hermes-function-calling-v1 and synthetic smolcode tool-use trajectories.

When to Use This Model

  • Agentic Coding Assistants: Ideal for developers building agentic coding assistants that require precise tool call formatting from small models.
  • Resource-Constrained Environments: Suitable for applications where larger models are impractical, leveraging a \u22642B parameter base model.
  • PowerShell Tooling: While the name includes "powershell", the core capability is general tool-calling, making it useful for any agentic task requiring structured tool output.

Finetunes
1 models