Fireworks AI launches f1, a new compound AI model specialized in complex reasoning tasks

Fireworks AI, a platform to build, deploy, and fine-tune generative AI models, has unveiled Fireworks f1, a compound AI model tailored for complex reasoning tasks. The f1 model is distinct in its integration of multiple open models at the inference layer, which Fireworks claims allows it to perform on par with or surpass many closed frontier models and leading open models in reasoning capabilities. It excels in coding, chat, math, and reasoning benchmarks.

The f1 model introduces a novel approach by utilizing prompting as a universal declarative programming language for generative AI applications. This design allows developers to focus on describing desired outcomes rather than detailing the methods to achieve them, simplifying the development process.

Alongside f1, Fireworks has released f1-mini, a more compact and faster variant. Both versions are currently accessible in preview with free access on the Fireworks AI Playground. Developers interested in leveraging f1's capabilities for complex reasoning use cases can apply for early access to the f1 API and design partnerships.