Cloudflare Launches Open-Weight Clef Models to Challenge TypeSafe's Jev

Cloudflare says Clef and Clef-flash outperform TypeSafe's Jev on several benchmarks. The company has positioned the models as an alternative in the emerging market for decision models, which are used to make structured choices rather than generate open-ended text. TypeSafe has not disclosed Jev's underlying architecture.

Clef uses a specially post-trained, frozen version of Qwen3.8-27B, while Clef-flash uses a frozen version of Qwen3.5-9B. During inference, the Qwen backbone performs a prefill-only pass, after which Clef scores choices in parallel. According to Cloudflare, Clef is fast and is faster than Jev.

Cloudflare tested Clef against Jev and other open decision models using the Jev Decision Index available on Hugging Face. Cloudflare's own ranking suggests Clef is slightly slower than other open models but more accurate, while Clef-flash is as accurate as most of the others and far faster. The scores were self-reported by Cloudflare and have not yet been reproduced for ranking on the official Decision Index.

Cloudflare also ran Clef against TypeSafe's own benchmarks and claimed it beat Jev in three out of four areas, losing only on agent trace observability. The company said Clef is not limited to classifying text and can also handle images and video. Clef supports a 64K context window. Jev can also handle up to 64K tokens across a request, although its state plus longest individual question is limited to 32K.