AI/News
Cloudflare's Clef-omni decision model takes audio and video as well as text and images, at $0.15 per million input tokens, and Clef-flash falls from $0.09 to $0.038
On Cloudflare's own tables Clef-omni beats TypeSafe's Jev on eight of ten public benchmarks but trails it on four of TypeSafe's five workflow tests and ties the fifth; the cheaper Clef-flash comes with a hosted context window cut from 64k to 24k tokens.
By Daily Aletheia · Checked against the primary source · 10 October 2026 · 2 min read
Cloudflare released Clef-omni on 9 October, a week after its first two open-weight decision models, Clef and Clef-flash. The new model is "able to take in audio (wav or mp3) and video (mp4 or webm) alongside text and images", and is built on "a Qwen3-Omni-30B-A3B-Instruct mixture-of-experts (MoE) foundation". The weights are on Hugging Face.
Speed, as Cloudflare states it: "Text-only decisions return in about 130 ms at the median, image inputs in about 150 ms, and audio clips in a few hundred milliseconds." And "a full 21-second video clip with sound is scored in about 1.5 seconds".
The prices: Clef-flash, "originally $0.09 per M input tokens, now $0.038 per M input tokens"; Clef stays at $0.24; Clef-omni launches at $0.15. Cloudflare says the cut makes Clef-flash "cheaper than Jev", TypeSafe's decision model, without giving Jev's price on the page. The same day, Microsoft priced its Microsoft-Decision-1 at $0.042 per million input tokens.
The trade for the cheaper Clef-flash: the hosted version "now has a context window of 24k, instead of 64k as previously advertised". Cloudflare says "only 0.24% of requests exceed 24k input tokens". The self-hosted weights are unchanged.
Clef itself is faster after serving moved to SGLang: median latency on about 800 tokens falls from 262 ms to 152 ms, on about 3,400 tokens from 616 ms to 305 ms, and on about 16,000 tokens from 2,721 ms to 1,635 ms.
Cloudflare's benchmark tables put Clef-omni ahead of Jev on eight of ten public tests and behind it on two, When2Call (63.3 against 80.97) and BRIGHT (42.0 against 47.52). On the five workflow tests from TypeSafe's own evaluations, Clef-omni trails Jev on four - invoice processing exact actions 60.2 against 61.8, for one - and ties on security incidents at 61.7.