A forgotten model ID in a fallback path can survive long after the primary route has been upgraded.
Cloudflare published a Workers AI deprecation notice on November 5, asking developers to remove references to 24 models before November 15. The list spans language, image, speech, vision, and embedding workloads. An application inventory needs to cover those secondary functions as well as its main chat model.
Sixteen language-model IDs are affected
The notice recommends newer frontier and smaller-model options. That is migration guidance, not evidence that a replacement accepts every old request or produces equivalent output.
@cf/deepseek-ai/deepseek-r1-distill-qwen-32b@cf/nvidia/nemotron-3-120b-a12b@cf/deepseek-ai/deepseek-v4-pro-0813@cf/deepseek-ai/deepseek-v4-flash-0731@cf/qwen/qwq-32b@cf/qwen/qwen3-30b-a3b-fp8@cf/moonshotai/kimi-k2.6@cf/moonshotai/kimi-k2.7-code@cf/zai-org/glm-5.2@cf/zai-org/glm-5.3-flash@cf/meta-llama/llama-2-7b-chat-hf-lora@cf/google/gemma-2b-it-lora@cf/google/gemma-7b-it-lora@cf/meta/llama-3.1-8b-instruct-fp8@cf/meta/llama-3.2-1b-instruct@cf/meta/llama-3.2-3b-instruct
Other tasks have named replacement suggestions
| Deprecated model | Cloudflare’s suggested replacement |
|---|---|
@cf/myshell-ai/melotts | @cf/deepgram/aura-2-en |
@cf/stabilityai/stable-diffusion-xl-base-1.0 | @cf/black-forest-labs/flux-2-klein-9b |
@cf/runwayml/stable-diffusion-v1-5-inpainting | @cf/black-forest-labs/flux-2-klein-9b |
@cf/bytedance/stable-diffusion-xl-lightning | @cf/black-forest-labs/flux-2-klein-9b |
@cf/lykon/dreamshaper-8-lcm | @cf/black-forest-labs/flux-2-klein-9b |
@cf/openai/whisper-tiny-en | @cf/deepgram/nova-3 |
@cf/llava-hf/llava-1.5-7b-hf | @cf/moondream/moondream3.1-9B-A2B |
@cf/google/embeddinggemma-300m | @cf/google/embeddinggemma-2 (available soon) |
The notice marks @cf/google/embeddinggemma-2 as available soon. Teams using the affected embedding model therefore need to verify actual availability rather than assume the replacement is already callable.
Trace the model ID through its consumers
Start with configuration and environment variables, then inspect background jobs, retry routes, and stored workflow definitions. A fallback that rarely runs can hide a deprecated model until another provider fails. Preserve the old-to-new mapping with the deployment record.
For embeddings, the migration also reaches stored vectors and search consumers. Compare output dimensions, normalization, and retrieval behavior before mixing vectors from different models. For speech and vision, inspect response formats and any downstream parsing. These are compatibility checks; we haven’t run them for this list.
Our AI Gateway alias article covers central model routing. An alias can simplify configuration changes, but it cannot make two models behaviorally identical. Which background or fallback route would be easiest to miss in your model inventory?
Abdessalam Alaoui
I'm Abdessalam Alaoui, founder of Musthave.AI. I write practical guides to AI models, tools, generative media.
