Diogo Almeida Releases Jev AI Model, Claiming Up to 200x Faster Performance

Summary

Diogo Almeida announced the release of Jev, a new frontier AI model that he says was developed during two years of work in stealth. In the same post, he described RLCD as a new method for training models and linked the project to his question of why increasingly capable chat models had not produced AGI. Almeida claims Jev is 20 to 200 times faster and 40 to 400 times cheaper, with output tokens available at no cost. He characterizes the model as providing “frontier composable intelligence” optimized for decisions. The post presents these figures and capabilities as the project’s own claims and does not provide benchmarks, methodology details, or independent validation.