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

TEXT GENERATIONConcurrent Unit Cost:1Model Size:3.1BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Jun 15, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

The seanpoyner/smolcode-coder-terraform-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 calls. This model addresses the limitation of smaller coder models by allowing them to drive agentic coding loops effectively. It is optimized for tool-use scenarios, particularly within the smolcode agentic coding assistant, by emitting byte-identical tool call formats for seamless integration.

Loading preview...

Overview

This model, seanpoyner/smolcode-coder-terraform-3b-tools, is a LoRA fine-tune of the Qwen2.5-Coder-1.5B-Instruct base model, featuring 3.1 billion parameters and a 32768-token context length. Its primary innovation is enabling the model to emit native <tool_call> function calls, a critical feature for driving agentic coding loops with smaller language models (SLMs).

Key Capabilities

  • Native Tool Calling: Unlike standard small Qwen-Coder models that output plain-text JSON for tool calls, this fine-tune generates the native <tool_call> format, which is directly parsed by runtimes like Ollama and llama.cpp.
  • Agentic Workflow Integration: Specifically built for smolcode, an SLM-optimized agentic coding assistant, to facilitate seamless tool-use.
  • Optimized Training: Trained using bf16 LoRA with assistant-only loss on attention and MLP projections, focusing loss specifically on tool calls and final answers. The training data includes NousResearch/hermes-function-calling-v1 and synthetic smolcode tool-use trajectories, all rendered through the exact apply_chat_template(tools=...) used at inference.
  • Improved Reliability (v2): Version 2 resolves a train/inference template mismatch from v1, ensuring high free-generation tool-call parse rates (≥90% on held-out prompts).

Good For

  • Developers building agentic coding assistants that require robust and native tool-calling capabilities from smaller language models.
  • Use cases where efficient and accurate function call generation is crucial for automating coding tasks.
  • Integrating with runtimes that expect the <tool_call> format for tool execution.

Finetunes
1 models