Meet Cloudflare Clef-omni: A 30B-A3B Multimodal Decision Model for Audio, Video, Image and Text

- Cloudflare’s Clef-omni is a 30B-A3B multimodal decision model that scores text, image, audio, and video in one call. It’s built on Qwen3-Omni, runs on Workers AI, and costs $0.15 per million tokens. Clef-flash dropped to $0.038, undercutting TypeSafe’s Jev. Hosted Clef is now 2x faster thanks to SGLang. Clef-omni leads intent classification but trails pure-text models on hard reasoning. It’s not a moonboy’s rocket fuel, but it’s efficient infrastructure. For the rest of you waiting for another L1 from nowhere, take notes: real utility beats vaporware every time. Stop trying to launch chains; start building pipelines that actually process data without melting your GPU. That’s how you survive the relay window.