TLDR Tech / DevOps
Powering the agents: Workers AI now runs large models, starting with Kimi K2.5
Cloudflare launched frontier open-source AI model Moonshot AI's Kimi K2.5 on its Workers AI platform, featuring a 256k context window and delivering 77% cost savings compared to mid-tier proprietary models in production use. The company reported that its internal security review agent, processing 7 billion tokens daily, caught over 15 confirmed issues while cutting annual costs from an estimated $2.4 million to a fraction of that amount, alongside rolling out new features like prefix caching discounts and a revamped asynchronous API to handle high-volume inference workloads.
Read the full story