Hammer2.0-1.5b: A Specialized Function-Calling Model

MadeAgents' Hammer2.0-1.5b is a lightweight 1.5 billion parameter model engineered for superior function-calling performance. It is built upon the robust Qwen 2.5 and Qwen 2.5 coder series, incorporating innovative function masking techniques during its training.

Key Capabilities

• Robust Function Calling: Specifically fine-tuned to accurately identify and execute functions based on user prompts.
• On-Device Agentic Applications: Designed to facilitate the development of intelligent, personalized applications that can run efficiently on edge devices.
• High Performance at Scale: Despite its compact size, the 1.5B model achieves notable performance on function calling benchmarks, often outperforming larger models in its category.
• Comprehensive Training: Trained on the extensive APIGen Function Calling Datasets (60,000 samples) and supplemented with the xlam-irrelevance-7.5k dataset, ensuring broad generalization.
• Stable and Robust: Evaluations on the Berkeley Function-Calling Leaderboard (BFCL-v3) and other academic benchmarks confirm its consistent and reliable performance.

Good For

• Developers creating agentic applications requiring precise tool and function invocation.
• Use cases demanding on-device processing where model size and efficiency are critical.
• Scenarios where accurate function parameter extraction and tool selection are paramount.
• Building applications that need to interact with external APIs or execute specific code functions based on natural language input.