sjan25/toolcalling-merged-demo
TEXT GENERATIONConcurrency Cost:1Model Size:2BQuant:BF16Ctx Length:32kPublished:Apr 9, 2026License:apache-2.0Architecture:Transformer Open Weights Cold
The sjan25/toolcalling-merged-demo is a 2 billion parameter Qwen3-based model, developed by sjan25, and fine-tuned using Unsloth and Huggingface's TRL library. This model is specifically optimized for tool-calling capabilities, leveraging its Qwen3 architecture and efficient fine-tuning process. It is designed for applications requiring robust function calling and interaction with external tools.
Loading preview...
Overview
The sjan25/toolcalling-merged-demo is a 2 billion parameter model developed by sjan25. It is based on the Qwen3 architecture and was fine-tuned from unsloth/Qwen3-1.7B-unsloth-bnb-4bit. The fine-tuning process utilized Unsloth for accelerated training, achieving 2x faster speeds, in conjunction with Huggingface's TRL library.
Key Capabilities
- Tool Calling: This model is specifically fine-tuned to excel at tool-calling tasks, enabling it to understand and generate calls to external functions or APIs.
- Efficient Training: Leverages Unsloth for faster and more efficient fine-tuning, making it a practical choice for developers looking for optimized models.
- Qwen3 Architecture: Built upon the Qwen3 foundation, providing a strong base for language understanding and generation.
Good For
- Function Calling Applications: Ideal for use cases where an LLM needs to interact with external tools, APIs, or execute specific functions based on user prompts.
- Agentic Workflows: Suitable for building AI agents that require robust tool-use capabilities to perform complex tasks.
- Developers Seeking Optimized Models: Benefits from the Unsloth fine-tuning, offering a performant model that was trained efficiently.