Custom reinforcement-learning environments and infrastructure for frontier labs and enterprises to train, evaluate, and optimize reliable agents.

A community benchmark for evaluating AI agents on realistic terminal tasks; Bespoke Labs is a core contributor.

An open collection of source-linked unsolved mathematics problems for researchers and models, with community result submissions.

An open-source library for scalable synthetic-data creation, structured LLM outputs, and dataset curation.

A collaborative open-data project that releases reasoning datasets and models; Bespoke Labs contributes alongside the wider research community.

An open-source reflective optimizer for prompts, code, and agents, developed with research collaborators.