Prime Inference 📈, agent population 📈, Claude academy 🎓

Prime Inference, a serving platform built on Prime Intellect’s GPU infrastructure, handles nearly a trillion tokens daily and supports large‑scale reinforcement learning rollouts, synthetic data generation, evaluations, and long‑running coding agents. The platform offers both serverless endpoints and reserved capacity across multiple data centers, enabling resilient serving of frontier open‑source models. Kolibri, an open‑weight English‑German Mixture‑of‑Experts model with 78.1 billion parameters, is available on Hugging Face under an Apache 2.0 license and supports tool calling and up to one million‑token contexts, targeting sovereign, mission‑critical work in government and regulated industries. Anthropic is investing $100 million in the Claude Frontier Academy to train 10,000 AI engineers by 2027, partnering with firms such as Accenture and Morgan Stanley to enhance AI fluency and enterprise technology integration. AI chips projected to ship through 2027 could run tens to hundreds of millions of concurrent frontier‑model agents, potentially supplying the equivalent of 140–720 million full‑time employees’ weekly hours; using 20 % of estimated central capacity would imply $2.6 trillion to $5.3 trillion a year in API‑equivalent spending. These developments highlight the rapid scaling of AI infrastructure, the strategic investment in engineering talent, and the economic implications of deploying large‑scale AI models across industries.