Cloudflare retires 24 Workers AI models on 15 November
A catalog refresh drops 24 older Workers AI models, from Stable Diffusion variants to small Llama and DeepSeek checkpoints, with no stated grace period after the cutoff.
Cloudflare is deprecating 24 older models on Workers AI, effective 15 November. The list spans image models (several Stable Diffusion variants), speech (whisper-tiny-en), vision (llava-1.5-7b-hf), an embedding model, and 16 LLMs including llama-3.2-1b-instruct, llama-3.2-3b-instruct, qwq-32b and deepseek-r1-distill-qwen-32b. Cloudflare says it is “refreshing the Workers AI model catalog to make room for new releases and updated models.”
For some models Cloudflare gives a direct swap, such as deepgram/nova-3 for whisper-tiny-en and black-forest-labs/flux-2-klein-9b for the Stable Diffusion variants. For the 16 LLMs it only gives general guidance: deepseek-v4.1-flash-1017 or glm-5.3 for frontier needs, gemma-4-26b-a4b-it or qwen3.8-27b for smaller models.
// before
model: '@cf/openai/whisper-tiny-en'
// after
model: '@cf/deepgram/nova-3'Why it matters: if any of these 24 model IDs are hardcoded in a Workers AI call, that request needs a new model string before 15 November. Cloudflare’s changelog doesn’t say what happens to calls made after the cutoff, just that references should be removed before then, so there’s no documented grace period to fall back on.
The caveat: the LLM replacements are general recommendations, not one-to-one swaps, so a model picked purely on price or size may need re-testing against the new options rather than a straight rename.