Prime Intellect Launches Inference API with 56 Models for Decentralized AI Evaluation

Image for Prime Intellect Launches Inference API with 56 Models for Decentralized AI Evaluation

San Francisco, CA – Prime Intellect, a company focused on democratizing AI development, has officially launched its Inference API, providing an OpenAI-compatible endpoint for evaluating environments on its Hub. The new API aims to simplify the process of assessing AI models, offering a unified billing system and a growing catalog of 56 models. This development is set to enhance the accessibility and efficiency of reinforcement learning evaluations within the decentralized AI ecosystem.

The company announced the launch via social media, stating, "> Our Inference API is live. Use it to evaluate environments on the Hub." They further detailed the API's features, including a "Simple prime env eval cmd," "One endpoint for all models," and a "Viewer for rewards and rollouts," emphasizing the ease of use and comprehensive evaluation capabilities. Prime Intellect, co-founded by Vincent Weisser and Johannes Hagemann in 2023, has been a proponent of open and decentralized AI, having previously released models like INTELLECT-2 and the Environments Hub for building and sharing reinforcement learning environments.

The Inference API offers token-based pricing, with charges for both input and output tokens, and integrates with Prime Intellect's existing unified billing system. This approach provides transparency and predictability for users managing their compute resources. The API is designed to route requests to various model providers, allowing for flexible model selection crucial for large-scale evaluations, and supports direct API access or usage through the Prime CLI.

Prime Intellect has raised $20.5 million in funding from investors including Founders Fund and CoinFund, underscoring investor confidence in its vision for decentralized AI. The company's platform aggregates global compute resources to facilitate distributed training and inference, aiming to make advanced AI development more accessible and less reliant on centralized infrastructure. This latest offering aligns with their broader mission to enable collective creation and ownership of open AI innovations.