Clef-flash
Cloudflare · Released 1 Oct 2026
Latency-focused sibling of Clef for faster, lower-cost decisions. Open-weight, Jev-API compatible, and hosted on Workers AI.
- Maker
- Cloudflare
- Release date
- 1 Oct 2026
- First seen
- 1 Oct 2026
- Access
- Workers AI; open weights on Hugging Face.
- Pricing (as publicly stated)
- As stated in Cloudflare's 9 Oct 2026 post: $0.038 per million input tokens, down from the earlier $0.09 rate. The hosted context window was reduced to 24k tokens as part of that price cut; Hugging Face weights were trained to support a larger window for self-hosting. See Cloudflare's developer docs for current rates.
- Modalities
- text, images
- Question types
- noul, choice, score
- Latency (as stated)
- In Cloudflare's 1 Oct 2026 launch benchmarks, Clef-flash's median latency was listed as 38.8 ms (p95 122.4 ms).
- Official announcement
- https://blog.cloudflare.com/clef-decision-models/
Benchmark claims
These are claims as published by the maker or in the linked official post. We have not independently rerun every benchmark.
Cloudflare published launch scores for Clef-flash alongside Clef and Jev, and later said Clef-flash remains performant at the lower price.
Names and marks belong to their owners. This page is an independent summary, not a product page from Cloudflare.