CLOUDFLARE LAUNCHES CLEF MULTIMODAL AI MODELS

The two new models—Clef (based on Qwen3.8-27B) and Clef-flash (based on Qwen3.5-9B)—handle images, video, and text processing. Both are available on Hugging Face for local deployment. Clef-flash prioritizes speed, while the larger Clef model targets superior performance. Cloudflare positions them as alternatives to existing options, though running either locally requires adequate hardware resources. The open-weight approach allows developers to download and run the models independently, rather than relying on API-based access. This gives users control over deployment and data handling, though it demands significant computational infrastructure. The release marks Cloudflare's expansion into generative AI model development, complementing its existing infrastructure services. Pricing details suggest higher costs than some alternatives, but Cloudflare emphasizes multimodal capabilities as a differentiator.