Cloudflare said on October 9 that hosted Clef-flash on Workers AI is getting cheaper, with a smaller context window. In the same Cloudflare announcement, the company also introduced Clef-omni, a new multimodal model for Workers AI and AI Gateway.
Hosted Clef-flash input pricing drops from $0.09 to $0.038 per million tokens. The tradeoff is a hosted context limit of 24K tokens, down from 64K. Cloudflare also says the self-hosted weights remain trained for 256K context, so the reduction applies to the managed version.
Cloudflare priced Clef-omni at $0.15 per million input tokens and kept Clef at $0.24. As of today, Clef-omni is available on Workers AI and through AI Gateway.
For teams optimizing around token cost, the cheaper hosted tier now leaves less room for long retrieval payloads, multi-document prompts, and accumulated agent memory.
