From question setting to scoring, everything is open-sourced end to end; the prime-rl training framework plus an environments hub ecosystem flywheel—this is what AI infrastructure should look like.

Crypto界网 news, Prime Intellect released Verifiers 0.2.0 and publicly unveiled a preview of the Verifiers v1 architecture. It is an open-source AI agent framework for generating, running, and scoring tasks, along with the open-source training framework prime-rl. Developers can download and deploy it, and share and obtain training environments through the environments hub and lab. The new Verifiers v1 splits task execution and agent runtime into three parts, improving reuse across environments. The 1.0.0 release plans to add multi-Agent environments and further enhance support for frameworks such as openenv, nemo-gym, and openreward.