seanpoyner/smolcode-coder-rust-1.5b-tools
TEXT GENERATIONConcurrent Unit Cost:1Model Size:1.5BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Jun 14, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold
seanpoyner/smolcode-coder-rust-1.5b-tools is a 1.5 billion parameter LoRA fine-tune of Qwen2.5-Coder-1.5B-Instruct, developed by seanpoyner. This model is specifically trained to emit native function calls, enabling agentic coding loops for small language models. It excels at integrating tool use within coding assistants by providing a byte-identical training target for tool calls.
Loading preview...
Model Overview
This model, smolcode-coder-rust-1.5b-tools, is a 1.5 billion parameter LoRA fine-tune of the Qwen2.5-Coder-1.5B-Instruct base model. Its primary innovation is the ability to emit native <tool_call> function calls, which is crucial for enabling agentic coding loops in small language models (SLMs).
Key Capabilities
- Native Tool Call Emission: Unlike base Qwen-Coder models that output tool calls as plain-text JSON, this fine-tune generates the
<tool_call>format directly parsed by runtimes like Ollama and llama.cpp. - Agentic Coding: Designed to drive agentic coding assistants, specifically built for the smolcode project.
- Optimized Training: Utilizes
bf16 LoRAwithassistant-only losson attention and MLP projections, focusing loss calculation exclusively on tool calls and final answers. - Data Strategy: Trained on a combination of
NousResearch/hermes-function-calling-v1for breadth and syntheticsmolcodetool-use trajectories for sharpness, ensuring the training target is byte-identical to the inference prompt.
Good For
- Developers building agentic coding assistants that require robust tool-use capabilities from small language models.
- Use cases where efficient and native tool call parsing is critical for integrating LLMs with external functions or APIs.
- Projects leveraging Qwen2.5-Coder-1.5B-Instruct and needing enhanced function-calling reliability.