Perplexity releases the classification model Decider, which it says costs about 1/12 as much as GLM-5.3-Flash run through an API, not counting local hardware

Concept illustration: a large board topped by a black band with a teal geometric shape and the lettering PERPLEXITY is split by a vertical line into two halves. On the left a black laptop lettered DECIDER sits beside a short stack of red discs and a tag reading COST, and on the right a black laptop lettered GLM-5.3-FLASH sits beside a much taller stack of grey discs and another COST tag; in the lower right corner a dark-haired person in dark clothing, seen from behind, faces the board. Printed on the image: About 1/12 Decider cost vs GLM-5.3-Flash via API, same requests, per Perplexity; excludes local hardware.

Perplexity released Decider v1.3 and the smaller Decider Flash v1.3 on November 12: given a passage and several questions, the models return the probability of each answer option for every question in one pass, without generating an answer word by word (Perplexity, 2026-11-12). Perplexity argues agents' small decisions can shift from text-generating LLMs to probability-only classifiers, backed so far only by its own tests.

Perplexity self-reports that on the same set of 1 million requests, Decider v1.3 cost about US$26.47, while GLM-5.3-Flash run through an API cost about US$312.21, about 12 times as much (our calculation). Decider ran locally, and Perplexity states that local hardware costs aren't included. The two sides aren't costed the same way.

The reference answers for accuracy also come from a large model. The Decisions API is Perplexity's external API for Decider. Perplexity compared results on Decisions API live traffic from before the launch: on the same questions, Decider v1.3's answers agreed with a frontier large model's answers 93.98% of the time, against 93.68% for the previous v1.2. That measures how closely Decider matches that large model, not whether Decider's answers are right.

Perplexity writes that the models are on Hugging Face under an Apache license. The Hugging Face model page is currently marked private and requires a login token, which doesn't match the announcement's statement that the weights can be downloaded.

Not proven yet: the reference answers for accuracy came from a frontier-model judge How we checked →

Next to watch: Decider's score on Hugging Face's decision-model leaderboard, and when the weights open for download.

Source and checks

This item ran in the brief of Nov 13, 2026, alongside 16 other items. How we checked it, the quote checks and what was still missing that day are on that issue's How we checked page.

Cite this item

SecondSource (2026). "Perplexity releases the classification model Decider, which it says costs about 1/12 as much as GLM-5.3-Flash run through an API, not counting local hardware." https://secondsource.io/en/news/2026-11-13-3/. Accessed access date.