Prime Intellect: FrontierBioML’s best agent scores 40.3/100

FrontierBioML’s launch post describes 25 computational biology tasks with 11 models, graded by running agents’ prediction scripts on held-out data. It lists Claude Mythos 5.2 first at 40.3, followed by Gemini 4 Carbon at 39.0 and Gemini 4 Argon at 36.8. Prime Intellect says the tasks are live on its Environments Hub.