Arch-Function-3B
| Property | Value |
|---|---|
| Model Size | 3B parameters |
| Base Architecture | Built on Qwen 2.5 |
| License | Katanemo license |
| Hugging Face | katanemo/Arch-Function-3B |
What is Arch-Function-3B?
Arch-Function-3B is a specialized large language model designed specifically for function calling tasks. As part of the Katanemo Arch-Function collection, it demonstrates impressive capabilities in understanding complex function signatures and producing accurate function call outputs from natural language prompts. The model achieves a remarkable 57.69% overall accuracy on the Berkeley Function-Calling Leaderboard (BFCL), positioning it competitively among leading models like GPT-4 and Claude.
Implementation Details
Built on the foundation of Qwen 2.5, Arch-Function-3B implements sophisticated mechanisms for parameter identification and function execution. The model requires transformers>=4.37.0 and is optimized for production environments with low-latency, high-throughput performance.
- Advanced function signature parsing and parameter matching
- Support for single, parallel, and multiple function calling
- Optimized for real-time API interactions
- Structured JSON output format for reliable function execution
Core Capabilities
- Single Function Calling: Execute individual function calls with high accuracy
- Parallel Function Calling: Process multiple instances of the same function with different parameters
- Multiple Function Calling: Handle complex queries requiring multiple different functions
- 85.19% accuracy in Single Turn scenarios
- 86.18% accuracy in Multi Turn interactions
Frequently Asked Questions
Q: What makes this model unique?
The model's specialization in function calling tasks sets it apart, with particular strength in parameter identification and API interaction. Its performance metrics are competitive with much larger models while maintaining efficient operation.
Q: What are the recommended use cases?
Arch-Function-3B is ideal for scenarios requiring automated API interaction, backend task automation, and complex function execution based on natural language inputs. It's particularly well-suited for production environments where reliable function calling is crucial.
