Summary
On October 9, 2026, Cloudflare added Clef-omni to Workers AI, a multimodal member of its Clef decision-model family that scores allowed answers across text, images, audio (WAV/MP3), and video (MP4/WebM) in a single pass at $0.15 per million input tokens. The update also cut Clef-flash input pricing from $0.090 to $0.038 per million tokens and made Clef up to 2x faster.
What changed
Cloudflare added @cf/cloudflare/clef-omni, a multimodal decision model that returns a calibrated probability for each allowed answer across text, images, audio, and video in one forward pass (64K context, $0.150 per million input tokens, ~20ms text / <100ms image-audio / ~300ms for a 21s video). It dropped Clef-flash input pricing from $0.090 to $0.038 per million tokens (shrinking its hosted context from 64K to 24K) and sped up Clef by up to 2x, partly via a move to SGLang, with weights unchanged.
Why it matters
Multimodal decision models let developers run classification, routing, scoring, and guardrails over audio and video without parsing generated text, and the steep Clef-flash price cut undercuts rival decision model Jev, signaling Cloudflare is competing aggressively on cost and latency for the agent control layer.
Evidence excerpt
Clef-omni accepts audio (WAV or MP3) and video (MP4 or WebM) alongside text and images, and does not generate text; it scores each allowed answer in a single pass. Clef-flash input pricing dropped from $0.090 to $0.038 per million tokens, and Clef is now up to 2x faster.