firefunction-v2

Introducing Firefunction-V2: Fireworks AI's Next-Generation LLM for Real-World Applications

Tal Peretz

Jul 15, 2024 — 2 min read

Fireworks AI is proud to announce the release of Firefunction-V2, our latest large language model (LLM) engineered to excel in real-world applications. Designed with a focus on function calling and multi-turn conversations, Firefunction-V2 brings remarkable advancements to the AI landscape.

Key Features and Capabilities

Function Calling

Firefunction-V2 is optimized for function calling, which enables it to interact with external APIs effectively. Unlike its predecessor, Firefunction-V2 supports parallel function calling, significantly enhancing its utility.

Performance

Our new model competes head-to-head with GPT-4o in function-calling tasks, scoring 0.81 on public benchmarks, slightly higher than GPT-4o’s 0.80. Impressively, Firefunction-V2 operates at 2.5 times the speed and 10% of the cost of GPT-4o, making it both faster and more cost-effective.

General Reasoning and Conversation

Firefunction-V2 balances function calling with general reasoning abilities, maintaining the conversational and instruction-following capabilities found in Llama 3.

Cost and Speed

Firefunction-V2 is highly cost-effective, priced at $0.9 per output token compared to GPT-4o’s $15 per token. It processes 180 tokens per second, compared to GPT-4o’s 69 tokens per second, ensuring rapid and affordable performance.

Integration and Accessibility

Accessible through Fireworks AI’s platform, Firefunction-V2 offers a speed-optimized setup with an OpenAI-compatible API, facilitating easy integration into existing systems. Users can also explore the model using a demo app and UI playground.

Development and Fine-Tuning

Firefunction-V2 was fine-tuned from the Llama3-70b-instruct base model using a curated dataset that included function calling and general conversation data. User feedback played a crucial role in its development, ensuring the model meets real-world application needs.

Practical Applications

Firefunction-V2 excels in multi-turn conversations and instruction following, making intelligent decisions about when to call functions and executing them accurately. It supports up to 30 function specifications, a significant improvement over its predecessor.

Market and User Reception

Fireworks AI has seen significant enterprise adoption, with companies like Uber, DoorDash, and AI-native startups building their stack around Fireworks. The developer community has responded positively, with beta testers praising its performance and cost-effectiveness.

Future Ambitions

Fireworks AI is committed to building the best platform for enterprises to deploy AI into production. We focus on latency, cost, quality, ownership, and developer experience. The team will continue iterating on Firefunction models based on user feedback and a commitment to practical solutions for developers.

Discover the future of AI with Firefunction-V2, and see how it can transform your real-world applications.

Unleashing Creativity with Stability.SD3-Large-V1:0 on Amazon Bedrock

The advent of Stability.SD3-Large-V1:0, a revolutionary text-to-image model, marks a significant milestone in the realm of generative AI. Integrated into Amazon Bedrock, this model is designed to transform how digital assets are created, offering unparalleled performance and quality in image generation. Model Architecture and Performance At the heart

Unlocking the Potential of Bedrock/Meta-Textgeneration-Llama-2-7B with AWS SageMaker

The Bedrock/Meta-Textgeneration-Llama-2-7B is a cutting-edge Large Language Model (LLM) developed by Meta as part of the Llama 2 series. This series ranges from 7 billion to 70 billion parameters, with the 7B model offering a balance of size and capability. Its development involved Supervised Fine-Tuning (SFT) and Reinforcement Learning

Exploring Azure AI's Phi-3-Medium-128K-Instruct: A Revolutionary LLM

Azure AI's Phi-3-Medium-128K-Instruct is a breakthrough in the field of language models, bringing forth a dense decoder-only Transformer model with 14 billion parameters. This model, part of Microsoft's Phi-3 family, stands out for its impressive performance across a wide range of benchmarks, making it a go-to

Unlocking the Potential of Azure AI's Phi-3.5-MoE-Instruct: A New Era in Artificial Intelligence

The Azure AI Phi-3.5-MoE-Instruct model, a groundbreaking addition to Microsoft's Phi-3.5 series, is revolutionizing the landscape of artificial intelligence with its innovative architecture and impressive capabilities. Model Architecture At the heart of the Phi-3.5-MoE-Instruct is the Mixture-of-Experts (MoE) model architecture, which includes 16 experts, each