The monthly · N° 0 · October 2026
Back to the issue
React
Share

From the daily edition · 12 Oct 2026Models and releases

Cloudflare released Clef-omni, which takes audio and video, and cut Clef-flash to $0.038 per million input tokens.

A week after releasing Clef and Clef-flash, Cloudflare released Clef-omni for audio, video, image and text input, and cut the price of Clef-flash while reducing its hosted context window.

206 wordsPublished 12 Oct 2026

It rests on 10 checked facts from 1 source.

Open on ainotis.com

Cloudflare published a post on 9 October 2026 about its Clef models, a week after it released Clef and Clef-flash. The new model, Clef-omni, takes audio (wav or mp3) and video (mp4 or webm) alongside text and images. Cloudflare says it built Clef-omni on a Qwen3-Omni-30B-A3B-Instruct mixture-of-experts model and released the weights on Hugging Face.

Cloudflare lists these prices per million input tokens: Clef-flash $0.038, down from $0.09; Clef $0.24, unchanged; and Clef-omni $0.15. To make the Clef-flash price possible, Cloudflare cut the hosted model's context window to 24k from the 64k it had advertised.

It says the weights on Hugging Face are untouched and were trained to support 256k tokens for self-hosting. Cloudflare says only 0.24% of requests exceeded 24k input tokens.

Cloudflare also says the hosted Clef is faster, with median speedups of 1.7 to 2.0 times depending on input size, from changes to its serving infrastructure rather than the weights. For Clef-omni it reports median latency of about 130 ms for text-only decisions and about 150 ms for images, which are its own figures.

How we checked this · the sources10 checked facts from 1 source.
  1. Introducing Clef-omni with full multimodality, plus a faster Clef and a cheaper Clef-flash · Cloudflare · archived copy

    1. Cloudflare's blog post introducing Clef-omni carries the publication date 2026-10-09.
      checked 12 Oct 2026
      "datePublished":"2026-10-09T18:27:34.129Z"
    2. Cloudflare says: "Today, we’re releasing Clef-omni, which takes in audio and video input alongside text and image." It names audio as wav or mp3 and video as mp4 or webm.
      checked 12 Oct 2026
      Today, we’re releasing Clef-omni, which takes in audio and video input alongside text and image. ... audio (wav or mp3) and video (mp4 or webm)
    3. Cloudflare says it released the Clef-omni weights on Hugging Face, and that it built the model on a Qwen3-Omni-30B-A3B-Instruct mixture-of-experts (MoE) foundation.
      checked 12 Oct 2026
      We also released the model open-weights on HuggingFace. ... We built this on a Qwen3-Omni-30B-A3B-Instruct mixture-of-experts (MoE) foundation
    4. Cloudflare lists: "Clef-flash – originally $0.09 per M input tokens, now $0.038 per M input tokens", "Clef – remains at $0.24 per M input tokens" and "Clef-omni – launched today at $0.15 per M input tokens".
      checked 12 Oct 2026
      originally $0.09 per M input tokens, now $0.038 per M input tokens ... remains at $0.24 per M input tokens ... launched today at $0.15 per M input tokens
    5. Cloudflare says: "The hosted version of Clef-flash now has a context window of 24k, instead of 64k as previously advertised."
      checked 12 Oct 2026
      The hosted version of Clef-flash now has a context window of 24k, instead of 64k as previously advertised.
    6. Cloudflare says the model weights on Hugging Face are untouched and "were trained to support a 256k context window should you choose to self-host it".
      checked 12 Oct 2026
      were trained to support a 256k context window should you choose to self-host it.
    7. Cloudflare says: "From our usage data, we see that only 0.24% of requests exceed 24k input tokens."
      checked 12 Oct 2026
      only 0.24% of requests exceed 24k input tokens
    8. Cloudflare reports median speedups for the hosted Clef of 1.7x for about 800 tokens, 2.0x for about 3,400 tokens and 1.7x for about 16,000 tokens, and says "most of the optimizations we made are at the serving infrastructure layer".
      checked 12 Oct 2026
      Median speedup 1.7× (~800 tokens), 2.0× (~3,400 tokens), 1.7× (~16,000 tokens) ... most of the optimizations we made are at the serving infrastructure layer
    9. Cloudflare says: "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."
      checked 12 Oct 2026
      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.
    10. Cloudflare says the post follows last week’s release of Clef and Clef-flash.
      Toned downchecked 12 Oct 2026
      Following last week’s release of Clef and Clef-flash

Next in this issue13 of 13The record

Mistral releases Large 4 weights, but paid hosting needs a separate licence

The monthly · N° 0 · October 2026 Back to the issue