AI Agent StorePerplexity · Provider documented
Perplexity Decider v1.2
A decision model for checking claims, grading records against a rubric or routing difficult requests. It returns probabilities and predefined choices, rather than a written answer.
Released 2026-10-16 · First reviewed 2026-10-17 · Checked 2026-11-03. Gateway dates its new offers October 16; the native changelog records the release in October without a day.
We still need comparable independent tests. You can compare the confirmed features and limits below, but we cannot yet give this version an Agent Fit or Practical Value rating. Test it on records with known answers, including cases that should go to a person.
What you can use today
Token prices measure the small pieces of text the model reads or produces. Capacity is the amount the service accepts in a request; check the limits below for any fallback to another model. A dash means the rate is unverified.
Exact provider API ID: pplx-decider-v1.2. Related questions can be answered together and share one input charge. Each question sees the others, which can change the answer. Score independent competing documents or candidates in separate requests. Confidence is calculated from the answer probabilities, not a separate guarantee of correctness.
Use pplx-decider-v1.2 for the current version or pplx-decider-latest to follow future updates. Revalidate your probability thresholds after a version change. Gateway also lists perplexity/pplx-decider-v1.2. Its zero output-limit placeholder is not a verified limit on answers. The native default is 10 requests per second per organization shared across both models, plus 15 million tokens per minute and 900 million per hour for this model. Higher limits require contacting Perplexity.
The native limit is fewer than 256K tokens for the shared information, images and each question. Large multi-question requests can be split. Up to 128 questions, 255 choice options, 10 score levels and a 32 MiB request body. Gateway’s 262,144 declaration is separate.
Output: Yes/no probabilities, choices and ordered scores. Compare decision models, limits and input costs ↗
Native input costs $0.02 per million tokens; answers are free and there is no per-request fee. Shared information is billed once within a group. If the service splits a large request into groups, it counts that information again. Include retries when comparing the total cost.
Provider release and specifications ↗
Check availability and plans ↗
Test it before switching
Give each model the same examples, tools and definition of success. Record correct outcomes, response time, retries and total cost. Check whether confidence scores match the answers you can verify.
Compare caching and long-context costs ↗