Prime Intellect

Prime Intellect provides an integrated stack for post-training open models and agents. Its offerings include the open-source Verifiers library and Prime CLI for building RL environments, an Environment Hub with community environments, hosted evaluations, hosted training, sandboxes for code execution, and an OpenAI-compatible inference API that serves GLM-5.3 and third-party models.

The site also lists on-demand GPU compute from one to 256 GPUs with hourly prices, reserved clusters sourced from multiple datacenter providers, and a spot market for idle GPUs. Its stated customers are teams that train their own models; a testimonial from Ramp describes training a small RL-tuned subagent with Prime Intellect. Revenue appears to come from compute rental, inference access and training services, with customers directed to book calls for dedicated capacity.

The company is headquartered in San Francisco, and its job listings also show New York, London and Berlin. The careers page lists 35 open roles, mostly in research engineering, applied research, infrastructure engineering, compute and finance, and growth. The site names Founders Fund, Radical, NVIDIA, Intel, Andrej Karpathy, John Schulman and Clem Delangue among its backers; no founding year or funding amounts are stated.

Product: Open RL environment framework (Verifiers), Prime CLI, hosted training and evaluations, OpenAI-compatible inference API, and on-demand GPU clusters.