Clef vs Clef-Flash
Cloudflare publishes two Clef decision models. They take the same input and return the same output format; they differ in size and price.
| Clef-Flash | Clef | |
|---|---|---|
| Parameters | 9B | 27B |
| Price (input) | $0.09 / M tokens | $0.24 / M tokens |
| Context window | 65,536 tokens | 65,536 tokens |
| Inputs | text, JSON, images, video | text, JSON, images, video |
| Question types | noul, choice, score | noul, choice, score |
| Model id | @cf/cloudflare/clef-flash | @cf/cloudflare/clef |
When to use Clef-Flash
Start here. It is the fast, cheaper option for high-volume decisions such as routing tickets, tagging content or screening images.
When to use Clef
Move a question to Clef when Clef-Flash’s probabilities are close together or wrong on your examples, or when the judgment is subtle. A common setup is to run Clef-Flash first and send only low-confidence cases to Clef.
Compare them yourself
Run the same question through both models in the free playground: write your question, run it with Clef-Flash, then switch the model to Clef and run again.
FAQ
Is Clef more accurate than Clef-Flash?
Clef is the larger model, so it is the one to try when Clef-Flash is unsure or wrong on your data. The best way to know is to run your own questions through both; the playground lets you switch models with one click.
Do Clef and Clef-Flash use the same API?
Yes. The request and response formats are identical. Change the model id in the URL and the model field.
Which one is cheaper?
Clef-Flash costs $0.09 per million input tokens; Clef costs $0.24, about 2.7 times more.
Facts from Cloudflare’s Workers AI model pages, checked 2026-10-02. We don’t publish benchmark claims we haven’t measured.