Benchmarks & research
Cloudflare's open-weight Clef-omni answers Jev-style typed questions about audio, video, images and text in one request, with a probability for every option.
Only you can see your notes.
Screenshot unavailable. Open the experiment ↗
Built on Qwen3-Omni-30B-A3B and released under Apache-2.0, it skips the usual transcribe-then-classify chain: Cloudflare says a 21-second video with sound is scored in about 1.5 seconds. On Workers AI it costs $0.15 per million input tokens, with audio and video counted as input tokens. Running the weights yourself takes about 64 GB of GPU memory.
Clef-omni suits checks that would otherwise need several tools, such as reviewing an installation from a photo, a recording of the machine running and a video of its fan. Ask a plain yes-or-no question about each, act on the confident answers and send unsure ones to a person. Cloudflare's own tests put it ahead of Jev on some tasks and behind on others, so try it on your own examples first.