What Perplexity announced
Perplexity Research published Perplexity Decider on November 12, 2026, announcing Decider v1.3 and Decider Flash v1.3 as successors to v1, v1.1, v1.2, and Flash v1.2. The post says Flash v1.3 is a smaller model distilled from Decider v1.3.
The blog and the official model cards say both models are available on Hugging Face under an Apache license, through OpenRouter, and through the Perplexity Decisions API. The official Hugging Face cards list license apache-2.0, name Qwen/Qwen3.8-27B as the v1.3 base model, and name Qwen/Qwen3.5-4B as the Flash base model, which the blog also describes as a distillation of the larger Qwen3.8-27B model.
The same official Hugging Face pages also state: "The repository is private: authenticate with a Hugging Face token that has access to it." This article therefore does not describe the weights as freely downloadable without access.
Who can use it
Perplexity says Decider supports binary and multi-class classification with ordered or unordered labels, returning probabilities that applications can use to classify, route, rank, and evaluate inputs. It describes three question types: Yes/No, multiple choice, and Score. New in v1.3 is locate, which extracts exact matching text plus start and end character locations, or an empty list if there is no match. The blog says the work is inspired by Jev, TypeSafe AI's System One Model.
Perplexity recommends v1.3 for nuanced work or cases where errors are costly, and Flash for high-volume routing, labeling, or filtering, and it advises tuning thresholds on a labeled development set.
The official Decisions API pricing page, checked November 14, 2026, lists pplx-decider-v1.3 at $0.02 per 1 million input tokens and pplx-decider-flash-v1.3 at $0.01, with output tokens free and no per-request fee. A documented example on that page says a request with 600 input tokens costs $0.000012 on v1.3 and $0.000006 on Flash.
Why it matters
If you already route or label text with a classifier, Decider v1.3 is Perplexity's current official decision model family, with a new locate question type and a cheaper Flash variant. It is a different product from Google's Gemini Decisions API, which AppFor.ai covered separately.
Perplexity also reports evaluation results in the same blog, including comparisons with Jev. Those figures are the vendor's claims. This article does not repeat them as independent scores, and AppFor.ai has not run the models.
Access and limitations
Access to the official Hugging Face repositories requires a token that is allowed to read them, even though the listed license is Apache 2.0. The v1.3 card lists a BF16 footprint of about 49 GiB, and Flash about 8.5 GiB. The announcement does not, by itself, change Perplexity's consumer search product.
This article is not a hands-on test and is not a comparison of Decider with the Gemini Decisions API beyond noting that they are different products.
Go to the source
Sources & further reading
Based on the official documentation below. Features and access can change; check the provider before making a decision.
- Perplexity Research: Perplexity Decider (November 12, 2026) (opens in a new tab)
- Hugging Face: perplexity-ai/pplx-decider-v1.3 (opens in a new tab)
- Hugging Face: perplexity-ai/pplx-decider-flash-v1.3 (opens in a new tab)
- Perplexity docs: Pricing (Decisions API rates, checked November 14, 2026) (opens in a new tab)
- Perplexity docs: Decisions API Quickstart (opens in a new tab)
Was this page useful?
Help us make our coverage more useful. This is feedback on the page, not a rating of the tool.
Go a little further