Prime Intellect launches serving platform for open models
Prime Intellect has introduced Prime Inference, a platform for serving frontier open-source models through serverless endpoints or reserved capacity on the company’s GPUs across multiple datacenters. Before its public release, the system handled nearly a trillion tokens per day in internal workloads including reinforcement-learning rollouts, synthetic-data generation, evaluations and long-running coding agents.