Cloudflare Launches Clef-omni, an AI Model for Open Weight Decision Making, Adds Support for Audio/Video InputCloudflare 推出开放权重决策 AI 模型 Clef-omni，新增支持音频 / 视频输入

IT House October 10th news, Cloudflare yesterday (October 9th) announced the launch of the open weight decision model Clef-omni, which supports a single API call to process text, images, audio and video. In terms of positioning, based on the previous Clef series model, the model extends to support multi-modal input, and the model weight has been opened up in Hugging Face. In terms of multimodal support, Clef previously supported text, images, and continuous images extracted from videos (video is extracted into a static image by time, and then these images are sequentially handed over to the model as input), while Clef-omni added audio and full video inputs, supporting wav, mp3, mp4, and webm formats. Cloudflare says developers don't need to set up separate speech transcription and audio and video splitting processes, and can process multiple inputs with one model. Cloudflare says Clef-omni is built on Qwen3-Omni-30B-A3B-Instruct and retains its primary understanding. The model is oriented towards structured decision tasks and does not generate regular text output. In the tests published by the company, the median response time for plain text requests is about 130 milliseconds and the image is about 150 milliseconds; the 21-second video with sound completes the score in about 1.5 seconds. Benchmark Clef-omni Clef-flash Jev BFCL: Exact matching rate 98.2 98.47 98.76 95.75 ToolRet: nDCG @ 10 66.6 69.19 66.43 65.28 API-Bank: Accuracy rate 92.7 91.93 93.11 88.19 Home appliances: Exact matching rate 69.3 82.95 97.73 52.27 When2Call: Accuracy rate 63.3 72.37 65.58 80.97 Banking77: Macro average F1 94.8 94.20 90.93 79.74 CLINC150 + OOS: Macro average F1 97.7 97.43 66.77 89.27 Bright: nDCG @ 10 42.0 45.91 39.26 47.52 Amazon ESCI: Macro average F1 57.8 57.48 57.39 55.21 PhishNChips: Accuracy rate 73.2 79.60 75.05 62.55 Cloudflare also reduced the price of Clef-flash input tokens from $0.09 (IT home note: the cash rate is approximately RMB 0.67) to $0.038 (the current exchange rate is approximately RMB 0.25) by approximately 58%. Clef-flash Hosted Context Window Sync reduced from 64k to 24k. Model input price output price Clef-flash $0.038 per million Tokens (current exchange rate about RMB 0.25) $0.12 per million Tokens (current exchange rate about RMB 0.8) Clef $0.24 per million Tokens (current exchange rate about RMB 1.6) $0.72 per million Tokens (current exchange rate about RMB 4.8) Clef-omni $0.15 per million Tokens (current exchange rate about RMB 1) $0.60 per million Tokens (current exchange rate about RMB 4)