Cloudflare is pruning the Workers AI model catalog to free space for newer releases. Effective November 15 2026, a defined list of large language models (LLMs) and other AI models will be deprecated, and any code, scripts, or configuration that still points to those identifiers will cease to function.
Scope of the Deprecation
The announcement groups the retiring assets into two categories. The LLMs slated for removal are:
@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
Additional non‑LLM assets are also being retired, each paired with a suggested successor:
@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
Recommended Replacement Models
Cloudflare points to four models that cover most use‑cases:
- Frontier workloads:
@cf/deepseek-ai/deepseek-v4.1-flash-1017and@cf/zai-org/glm-5.3 - Smaller footprints:
@cf/google/gemma-4-26b-a4b-itand@cf/qwen/qwen3.8-27b
These replacements are positioned as more token‑efficient and higher‑throughput options. Pricing details remain on the Workers AI pricing page.
Operational Impact
Any deployment that references a deprecated identifier will encounter runtime failures once the catalog is refreshed. Practically, this means:
- CI/CD pipelines must be audited for hard‑coded model strings and updated to the new
@cf/…identifiers. - Configuration files, environment variables, and secret stores that hold model names need revision.
- Automated tests that validate model loading should be extended to cover the new models, ensuring compatibility with existing prompt formats and input schemas.
- Because the recommended models claim better token efficiency, engineers may observe reduced request latency or lower compute cost after migration, but they should verify these claims against their own workloads.
- Pricing changes are possible; teams should review the Workers AI pricing page to anticipate cost differences before promoting the new models to production.
Related CloudNinjas coverage: AI engineering.
What This Means For Practitioners
Actionable steps:
- Inventory every reference to a Workers AI model across repositories, deployment manifests, and infrastructure‑as‑code definitions.
- Cross‑check each entry against the deprecation list; replace any match with the suggested alternative or one of the four recommended models.
- Run a smoke test against the new model identifiers in a staging environment to confirm prompt compatibility and performance expectations.
- Update monitoring dashboards to track model‑specific metrics (e.g., token usage, latency) for the new models.
- Document the change in change‑management records and communicate the timeline (deadline: November 15 2026) to all stakeholders.
By completing these steps before the cutoff date, teams avoid service disruption, maintain expected performance, and stay aligned with Cloudflare’s evolving AI offering.
