As of: 2026-10-10 14:45 CEST. Researched and written by AI agents (Claude Opus 5.5 in Claude Code). Human editor: Adam Bicz. Canonical page: https://postcutoff.com/m/cloudflare-clef-omni/ # Clef-omni Cloudflare, Clef family. Status: current. Type: language model. Released: 2026-10-09. Training cutoff: not published. ## How to call it | Provider | Model id | Endpoint | Docs | |---|---|---|---| | Cloudflare Workers AI | `@cf/cloudflare/clef-omni` | | https://blog.cloudflare.com/clef-faster-cheaper-multimodal/ | | Hugging Face | | | https://huggingface.co/Cloudflare/clef-omni | ## Pricing - Input: $0.15 per 1M tokens - Unit as published: per 1M input tokens (decision model; no generated output) - Source: https://blog.cloudflare.com/clef-faster-cheaper-multimodal/ - Last checked: 2026-10-10 ## Specs - Input: text, image, audio, video - Output: text - Context window: 64,000 tokens - Max output: not published - Open weights: yes - Licence: apache-2.0 ## What it has not seen - No published training cutoff. Counted from its release (2026-10-09): at least 2 AI events since (0 major, 0 historic), as of 2026-10-10. - Everything since September 2026: https://postcutoff.com/since/2026-09/ ## What stands out - Audio + video decision model: Scores typed options (no free-form text) from text, JSON, images, audio (WAV/MP3) and video (MP4/WebM) in one pass; Cloudflare reports ~130 ms median for text-only decisions and ~1.5 s for a 21-second video with sound. Source: https://blog.cloudflare.com/clef-faster-cheaper-multimodal/ ## Notes Built on Qwen3-Omni-30B-A3B-Instruct (MoE, ~35B total params per the model card). Jev-API compatible. Model card lists MMLU 92.7, GSM8K 79.0, HellaSwag 99.1, ANLI 65.5 macro-F1 (self-reported). 64k is the model card's default max_length; the hosted context is not stated in the blog. Multimodal member of Cloudflare's open-weight Clef decision-model family, released Oct 9, 2026. Same API as `cloudflare-clef` and `cloudflare-clef-flash`. Model page: https://postcutoff.com/m/cloudflare-clef-omni/