Cloudflare tries to outplay Jev with open-weight Clef models
Cloudflare has released two decision models called Clef and Clef-flash, just two weeks after TypeSafe's Jev model generated significant industry attention. The company claims its models are faster and more capable than Jev whilst also being open-weight and runnable locally, directly challenging what appeared to be an emerging market leader. This matters because decision models are becoming critical infrastructure for automating structured business decisions at scale.
Clef uses Qwen's large language models as its backbone and can process images and video alongside text, whereas Jev is limited to text. According to Cloudflare's benchmarks, Clef outperforms Jev in three of four test areas, though these results have yet to be independently verified on the official Decision Index. Whilst Clef costs $0.24 per million tokens compared to Jev's $0.042/M (six times more expensive), it offers a 64k context window and can be downloaded from Hugging Face for local deployment, requiring 41GB of VRAM for Clef-flash and 85GB for the full Clef model.
- Cloudflare releases Clef models claiming to outperform Jev after just two weeks
- Can handle images and video alongside text, but costs six times more
- Available for local deployment via Hugging Face or hosted on Cloudflare's platform
New here? Start with this
Decision models are artificial intelligence systems designed to make routine business decisions automatically – approving loan applications, assigning customer support requests, or assessing credit risk. Organisations rely on such systems to handle high volumes of transactions with minimal human involvement.
TypeSafe provides Jev, a decision model focused specifically on automating structured business decisions. Cloudflare has released competing models called Clef and Clef-flash, signalling that decision models are becoming viewed as critical infrastructure by major technology companies.
The emergence of multiple competitors in decision models reflects growing organisational demand for automated decision-making systems. This matters because AI systems increasingly make consequential decisions affecting people and businesses, making the quality and reliability of these models a significant concern.