Fireworks AI Releases Firefunction-v2: An Open Weights Function Calling Model with Function Calling Capability on Par with GPT4o at 2.5x the Speed and 10% of the Cost
Introduction to Firefunction-v2
Firefunction-v2 is an open-source function-calling model designed for real-world applications, integrating multi-turn conversations, instruction following, and parallel function calling. It offers a robust and cost-effective solution with superior speed and functionality compared to high-end models like GPT-4o.
The Creation Process
Driven by user feedback, Firefunction-v2 balances function calling and general reasoning abilities, fine-tuned from the Llama3-70b-instruct base model. It maintains broad capabilities while enhancing function-calling performance.
Evaluation and Performance
Firefunction-v2 outperforms its predecessor and other models in function-calling tasks, achieving higher scores in parallel function calling and multi-turn instruction following, demonstrating its adaptability and intelligence in handling complex tasks.
Highlighted Capabilities
Firefunction-v2 reliably supports up to 30 function specifications, significantly improving over previous versions, crucial for handling multiple API calls efficiently and excelling in instruction-following.
Getting Started with Firefunction-v2
Accessible through Fireworks AI’s platform, Firefunction-v2 offers a speed-optimized setup with an OpenAI-compatible API, allowing integration into existing systems with minimal changes. Users can explore the model through a demo app and UI playground.
Conclusion
Firefunction-v2 sets a new standard for real-world AI applications, balancing speed, cost, and performance. It has the potential to revolutionize how function calls are integrated into AI systems, driven by user feedback and a dedication to providing practical solutions for developers.
