Arch-Agent-7B: Advanced Function Calling for Agentic Workflows

Arch-Agent-7B is a 7.6 billion parameter model developed by katanemo, part of the Arch-Agent series, specifically engineered for robust function calling and agent-based applications. It is designed to manage complex, multi-step tasks by intelligently selecting tools, planning adaptively, and integrating seamlessly with external APIs.

Key Capabilities

• Multi-Turn Function Calling: Maintains context across dialogues for natural, evolving tool use.
• Multi-Step Function Calling: Plans and executes sequences of function calls, dynamically adapting to intermediate results and decomposing goals.
• Agentic Capabilities: Provides advanced decision-making and workflow management for complex tasks, including tool coordination and error recovery.

Performance

Arch-Agent-7B demonstrates leading performance on the Berkeley Function-Calling Leaderboard (BFCL), showcasing its proficiency in complex function-calling scenarios. The model is evaluated with a context length of 64K, utilizing YaRN scaling for multi-turn evaluations.

Good For

• Developers building sophisticated AI agents requiring reliable function calling.
• Applications needing to manage multi-step and multi-turn interactions with external tools.
• Scenarios demanding precise and adaptive planning for complex tasks.