Cloudflare’s bigger, better, faster AI platform

Cloudflare announces major Birthday Week 2024 upgrades across its AI platform: Workers AI gains more powerful, globally distributed GPUs enabling faster inference and larger models (including Meta Llama 3.1 70B and the 3.2 family), a new Run Any* Model closed beta with static/dynamic catalogs, and a shift from “neurons” to transparent unit-based pricing and refreshed free tiers. AI Gateway adds persistent logs with annotations, cost/duration insights, Logpush export, and open-beta human-in-the-loop evaluations with cost analysis. Vectorize v2 reaches GA with up to 5M-vector indexes, 95% lower median query latency, up to 100 results per query, and substantial price cuts (75% query, 98% storage), plus a new free tier. The roadmap emphasizes tighter integration among Workers AI, AI Gateway, and Vectorize for end-to-end observability, evaluation, and low-latency RAG experiences.