Service eligibility and regional restrictions
PuppyIP serves only compliant overseas businesses and their authorized personnel. Proxy services are not available in mainland China. The service may only be used for lawful business activities outside mainland China. Use of this service within mainland China is prohibited.
Hosting a proxy IP or server overseas does not change these restrictions. The service must not be provided to end users in mainland China through relaying, forwarding, sharing or resale. Before use, read the Terms of Service.
Key Takeaways
- Cloudflare's November 5 notice plans to deprecate 16 LLM IDs and eight other model IDs on November 15.
- The notice groups four suggested LLM targets by use case; it does not provide a one-to-one replacement table or promise compatibility.
- Eight other models have suggested targets. embeddinggemma-2 is marked available soon, which does not establish current access.
- The notice gives no cutoff time or timezone, account-specific access status, failure behavior, automatic forwarding or fixed price.
Plan around the announced date
On November 5, 2026, Cloudflare published a planned Workers AI model deprecation notice. It asks apps to remove the IDs below before November 15, 2026. This article was checked on November 6; the notice describes a plan and does not show that these models are already unreachable.
These IDs identify Cloudflare Workers AI hosted endpoints. The notice does not establish retirement of the same models at their original vendors or on other platforms, and it does not list real-time availability for each account and region.
Check all 16 affected LLM IDs
These 16 full IDs are transcribed from Cloudflare's notice. Search for each exact string in code and configuration:
@cf/deepseek-ai/deepseek-r1-distill-qwen-32b, @cf/nvidia/nemotron-3-120b-a12b, @cf/deepseek-ai/deepseek-v4-pro-0813, and @cf/deepseek-ai/deepseek-v4-flash-0731.
@cf/qwen/qwq-32b, @cf/qwen/qwen3-30b-a3b-fp8, @cf/moonshotai/kimi-k2.6, and @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, and @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, and @cf/meta/llama-3.2-3b-instruct.
Treat the LLM targets as grouped suggestions
For frontier model needs, the notice recommends @cf/deepseek-ai/deepseek-v4.1-flash-1017 and @cf/zai-org/glm-5.3. For smaller model needs, it recommends @cf/google/gemma-4-26b-a4b-it and @cf/qwen/qwen3.8-27b. These targets are grouped by need; the notice does not map each old ID to a unique successor.
The recommendations do not promise compatible API schemas, tool calls, output formats, quality, prices or account access. Validate candidates against the inputs and dependencies of the application that will use them.
Review the other eight models and suggested targets
The suggested target for @cf/myshell-ai/melotts is @cf/deepgram/aura-2-en. The suggested target for @cf/openai/whisper-tiny-en is @cf/deepgram/nova-3.
Four image-generation IDs each map to @cf/black-forest-labs/flux-2-klein-9b: @cf/stabilityai/stable-diffusion-xl-base-1.0, @cf/runwayml/stable-diffusion-v1-5-inpainting, @cf/bytedance/stable-diffusion-xl-lightning and @cf/lykon/dreamshaper-8-lcm.
The suggested target for @cf/llava-hf/llava-1.5-7b-hf is @cf/moondream/moondream3.1-9B-A2B. For @cf/google/embeddinggemma-300m, the notice suggests @cf/google/embeddinggemma-2 and labels it available soon. That label does not prove it is currently available to your account.
Inventory every path that can select a model
Check each app's Workers configuration and environment variables, bindings, REST requests, SDK wrappers, model routers or Gateways, background queues, fallback rules and deployment manifests. Search for the 24 full IDs, then confirm the ID selected in each runtime; non-default settings and fallback paths can hold separate references.
For each match, record the app and environment, call path, current ID, candidate to evaluate, owner and planned switch date. Do not put credentials or customer inputs in the inventory. This is a practical migration step, not a separate configuration requirement published by Cloudflare.
Validate, then migrate in stages
In an authorized, isolated test environment, use nonsensitive representative requests to check input and response structures, downstream parsing, any tools or streaming paths the app uses, task quality, errors and latency. Estimate cost from Cloudflare's current official pricing material and the terms that apply to your account; do not infer price, performance or access eligibility from a model name.
Switch one call path or a small amount of traffic first, review errors and results, then expand. Keep the previous configuration recoverable. Use a rollback target only if it is still available and your account is authorized to call it. If the response shape or results do not fit, pause the rollout and continue validation instead of moving all traffic at once.
Limits the notice leaves open
The notice says to update apps before November 15, 2026, but gives no time of day or timezone. It does not say which HTTP error old IDs will return, whether requests will be forwarded automatically, whether each account can use the suggested targets or what fixed prices apply. Do not describe planned deprecation as a failure that has already happened or assume automatic compatibility or migration.
If a candidate is unavailable or not authorized for your account, check Cloudflare's current documentation and account status and adjust the rollout plan. Do not bypass eligibility or pricing restrictions to complete a migration.
Sources
Frequently Asked Questions
Must I change models if my app uses none of the listed IDs?
Check the exact strings in code, configuration and fallback paths first. If there are no matches, this notice alone does not establish a need to replace your model. Follow the relevant lifecycle notices for IDs outside the list.
Are the same models being retired by their original vendors or on other platforms?
This notice covers the listed Cloudflare Workers AI hosted IDs. It does not establish matching retirements at original vendors or other platforms. For another endpoint, check that provider's own lifecycle notice.
Can I keep an old ID as a fallback until its suggested target becomes available?
A fallback configuration does not extend the announced deadline. The notice still asks apps to remove these references before November 15. Retaining historical configuration does not guarantee that an old model will remain callable. Confirm another available, authorized target or prepare a degraded service mode.