Cloudflare, best known as web infrastructure, is now launching AI models of its own. Michelle Chen, who runs the company's AI platform, announced at the AI Engineer Code Summit that Cloudflare considers itself a 'neolab' — a term coined by The Information for a new generation of research-focused AI startups outside OpenAI and Anthropic.
Two system one models were announced. Clef is a decision model inspired by TypeSafe AI's Jev model, while Clairic is what Cloudflare calls an 'extract' model, a new type of system one model. Both tie into the company's infrastructure business.
More is on the way. Chen said in an interview with Latent Space that Cloudflare is experimenting with reasoning models and other LLMs as well.
- The focus is on post-training, not pre-training from scratch; Chen said the team can take existing models and make them more efficient and effective.
- Frontier-scale pre-trained models are explicitly not the goal.
- Chen suggested Cloudflare's infrastructure makes system one models a natural fit, but that the door is now open to other model types.
For developers, this signals that Cloudflare may start bundling its own models into the platform it already runs, rather than acting purely as a neutral infrastructure layer.
Summary by angular.love. The original source has the full details.