Cloudflare’s AI Search service received a significant update on August 26 2026. The platform now supports six new Workers AI text‑generation models, broadening the range of large‑language‑model options available to developers without requiring extra provider credentials.
New models and their context windows
- @cf/deepseek‑ai/deepseek‑v4‑flash‑0731 – 1,048,576 tokens
- @cf/deepseek‑ai/deepseek‑v4‑pro‑0813 – 1,048,576 tokens
- @cf/openai/gpt‑oss‑120b – 128,000 tokens
- @cf/openai/gpt‑oss‑20b – 128,000 tokens
- @cf/qwen/qwen3.8‑27b – 262,144 tokens
- @cf/moonshotai/kimi‑k2.7‑code – 262,144 tokens
All six models run on Cloudflare’s Workers AI runtime, which means they inherit the platform’s built‑in security, scaling, and latency benefits. Because they are hosted internally, developers do not need to supply an external provider API key to use them.
How to select a model
When creating or updating an AI Search instance—either through the Cloudflare dashboard or via the public API—you can now choose any of the new models. The selection process is identical to existing models; the only change is the expanded list presented in the UI or API payload.
Where to find the full list
The complete, up‑to‑date catalog of supported Workers AI models is available on Cloudflare’s changelog page and the corresponding Supported models documentation.
This expansion reinforces Cloudflare’s strategy of integrating generative AI directly into its edge network, offering developers a richer set of tools for search, summarisation, and content generation without the complexity of managing external API keys.