CLEFBENCH / RESEARCH NOTE / 2026-10-08 / UPDATED 2026-10-10
Clef vs Perplexity Decider: APIs, limits and pricing
Compare documented question types, input limits and direct API prices. Clef is tested on ClefBench; Perplexity Decider is not.
Clef and Perplexity Decider accept context and typed questions rather than an open-ended chat prompt. This page compares their documented interfaces. We have not run Perplexity Decider in ClefBench's benchmark or playground, so there is no measured accuracy or latency ranking between these two here.
Official interfaces and limits
| Dimension | Cloudflare Clef | Perplexity Decider v1.1 |
|---|---|---|
| Direct model ID | @cf/cloudflare/clef | pplx-decider-v1.1-27b |
| Access | Workers AI binding or account REST endpoint | POST https://api.perplexity.ai/v1/decisions |
| Decisions | noul, choice, score | noul, choice, score |
| Official question count | 1–64 | 1–128 |
| Published input constraint | 65,536-token directory context, with documented long-state truncation | Input must be below 262,144 tokens, including state, images and questions |
| Available on ClefBench | Yes, text/JSON only | No |
| ClefBench measurements | 60-case evaluation | Not tested |
Check the Clef hosting limitations before comparing context numbers. Perplexity additionally documents 1–255 choice options, 1–10 score levels and a 32 MiB request body. Limits belong to the named direct interfaces; they are not ClefBench allowances.
What the response means
Both interfaces organize results by question name. A yes/no answer is a probability of yes. A choice contains an option and distribution; an ordered score is a weighted level index. Perplexity's confidence is not the highest option probability and must not be treated as a cross-provider accuracy percentage.
Keep the same state, question wording and allowed options when comparing APIs. Save the returned model identifier, usage and raw result. Different question definitions make the outputs different experiments even when the surrounding text is identical.
Official input-token prices
| Model / source | USD / 1M input tokens |
|---|---|
| Clef | $0.240 |
| Perplexity Decider v1.1 | $0.020 |
These are model input rates, not a performance verdict. Perplexity documents input billing for state, questions and images, with no output or per-request fee. Cloudflare account allowances and plan costs are separate considerations.
Our official API cost estimate includes both rates and explains its assumptions. ClefBench subscription pricing covers this site's available models and does not include Perplexity Decider access.
How to decide without an unsupported ranking
Choose a labeled sample representing your actual decisions. Evaluate each direct interface under the same definitions; record incorrect choices, missed positive cases, rating error, response time and billed input. For long or image-heavy inputs, verify what was accepted and processed before comparing accuracy.
If you want an immediate browser trial, try Clef here. If Perplexity's documented limits fit your workload, follow its official quickstart and test it in your own account. We cannot infer which is better for your task from token prices alone.
Sources checked October 10, 2026: Cloudflare model card, Perplexity Decisions quickstart, Perplexity pricing.