Clef-omni
Cloudflare's mixture-of-experts decision model that reads text, JSON, images, audio, and video in one request and returns a probability for every allowed answer.
Overview
Cloudflare shipped Clef-omni on October 9, 2026, the mixture-of-experts member of its open Clef decision-model family, fine-tuned from Qwen3-Omni-30B-A3B (30B total, 3B active parameters) and served on Workers AI. It turns a state plus a schema of typed questions into decisions, reading the state directly as text, JSON, images, audio (WAV/MP3), or video (MP4/WebM) and returning a probability for every allowed option of every question, rather than generating free text. The same changelog update also cut Clef-flash's price and sped up the base Clef model.
This is a sourced directory entry. Features and capabilities are reported in the linked sources; this entry does not document independent Aymorphic testing.
Key features
- Reads text, JSON, images, audio (WAV/MP3), and video (MP4/WebM) in a single request
- Mixture-of-experts fine-tune of Qwen3-Omni-30B-A3B (30B total / 3B active parameters)
- Returns a calibrated probability per allowed option for each typed question, not generated text
- 65,536-token context window; images cost 64-1,024 input tokens each
- Released alongside a Clef-flash price cut and Clef speed improvements
Use cases
- Multimodal agent decision steps (e.g. triage a support ticket with an attached photo or voice note)
- Structured moderation/classification over mixed text, image, audio, and video inputs
- Workers AI-hosted guardrail or routing layer for agentic applications
Pricing
$0.15 per million input tokens, $0 per million output tokens, via Cloudflare Workers AI.
Pricing reflects the source record and may change. Confirm current terms with the provider.
Dates and sources
- Released
- 9 Oct 2026
- Record updated
- 12 Oct 2026
- Record last checked
- 12 Oct 2026