FireFunction-v1
| Property | Value |
|---|---|
| Parameter Count | 46.7B |
| Model Type | Function Calling |
| License | Apache 2.0 |
| Tensor Type | FP16 |
What is firefunction-v1?
FireFunction-v1 is a state-of-the-art function calling model developed by Fireworks AI, designed to achieve near GPT-4 level quality for structured information generation and routing decision-making. This model stands out for its exceptional speed, operating approximately 4x faster than GPT-4 when deployed on the Fireworks platform.
Implementation Details
The model is built on a transformer architecture and optimized for single-turn request routing and structured information extraction. It features API compatibility with OpenAI's function calling interface and supports up to 20 function specifications in its function pool.
- 46.7B parameter model with FP16 precision
- Compatible with OpenAI function calling API
- Unique support for "any" parameter in tool_choice
- Optimized for high-speed inference
Core Capabilities
- Single-turn request routing to functions
- Structured information extraction
- Fast inference processing
- Function selection from large function pools
- API compatibility with existing frameworks
Frequently Asked Questions
Q: What makes this model unique?
FireFunction-v1 stands out due to its combination of GPT-4 level quality, significantly faster inference speeds, and unique support for the "any" parameter in tool_choice. It's also one of the few models specifically optimized for function calling with a commercial-friendly license.
Q: What are the recommended use cases?
The model excels in two primary areas: single-turn request routing when choosing from up to 20 function specifications, and structured information extraction. It's not recommended for general multi-turn chat or parallel/nested function calls in a single response.
