Understand this story
On November 12, 2026, Perplexity released two decision models, Decider v1.3 and Decider Flash v1.3, adding verbatim text span extraction capability and opening weight downloads under the Apache-2.0 license. According to a Unite.AI report, the company said the larger model achieves an average accuracy of 91.6% across 63 query classification tasks, 6.54 percentage points higher than its production classifier; the new Locate feature returns matching text along with start and end character positions, and can be used for named entity recognition, personally identifiable information detection, and extractive question answering, returning an empty list when there is no match, and both models can also be invoked through OpenRouter and Perplexity's Decisions API. A subsequent Artificial Analysis report said the two models scored 70 and 64 respectively on the Artificial Analysis Decision Index v2.1, with the former holding 6th place and the latter rising to 10th; their parameter sizes are 27B and 4B respectively, their context windows are 256K tokens, and their input prices are $0.02 and $0.01 per 1M tokens respectively.
AI-generated from public reports
Latest development · Report
Perplexity Releases Decider v1.3 and Flash v1.3 Decision ModelsDevelopments
1 confirmed developments
- Source:Unite.AI
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Nov 14
Artificial Analysis
Perplexity Releases Decider v1.3 and Flash v1.3 Decision Models
Perplexity has released Decider v1.3 and Decider Flash v1.3, which score 70 and 64 respectively on the Artificial Analysis Decision Index v2.1; the former holds 6th place, while the latter rises to 10th. Decider v1.3 ranks 4th on AA-Paperclip with a score of 69 and 3rd on EvalSafe-O*NET-AA with a score of 89; Flash v1.3 improves by 4 points over v1.2, with AA-Symmetry and When2Call each improving by 12 points. The two models have parameter sizes of 27B and 4B respectively, a context window of 256K tokens, input prices of $0.02 and $0.01 per 1M tokens respectively, and are released under the Apache License 2.0.
MarkTechPostRelated discussion
MarkTechPost joined this discussion
This post contributes to heat; it is not a formal report or confirmed fact.
Nov 13
Aravind Srinivas
Perplexity Open-Sources Decision Models, an Agent Security Model
Perplexity CEO Aravind Srinivas announced the open-sourcing of Decision Models, saying they can process billions of tokens of telemetry data, identify security issues, and help fix them, at a scale and cost structure that was previously difficult to achieve. The models are free on OpenRouter for one week. The quoted tweet notes that Perplexity internally used Decider to read 1 billion tokens across 7,000 agent sessions in a single day, and the issues found led to 14 fixes to its agent security rules. Both models are Apache-licensed and free on OpenRouter until November 19.
Aravind Srinivas
Perplexity Open-Sources Decider v1.3 and Decider Flash v1.3 Decision Models
Perplexity has open-sourced two decision models, Decider v1.3 and Decider Flash v1.3, and is offering them free on OpenRouter for one week. v1.3 adds a locate capability that returns matching text and character offsets based on a description, suitable for batch decision-making as well as evidence and entity extraction. The company says it significantly outperforms the best general-purpose extraction models on 17 out-of-domain NER, PII, and extractive QA tasks, and achieves 6.54 percentage points higher accuracy than the currently used classifier on its own 63 production classification tasks. The weights have been released on Hugging Face.
Unite.AI
Perplexity Releases Decider v1.3 and Decider Flash v1.3 Models, Adding Verbatim Span Extraction
On November 12, Perplexity released Decider v1.3 and Decider Flash v1.3, adding verbatim text span extraction capability and releasing the weights for download under the Apache-2.0 license. The company says the larger model achieves an average accuracy of 91.6% across 63 query classification tasks, 6.54 percentage points higher than its production classifier. The new Locate feature returns the matched text along with its start and end character positions, and can be used for named entity recognition, personally identifiable information detection, and extractive question answering; it returns an empty list when there is no match. Both models can also be called via OpenRouter and Perplexity's Decisions API.
Heat trend for this story
- Current
- 7
- Period peak
- 22 Nov 14 01:00
- 24-hour change
- -65%
Move the pointer or select the chart to inspect hourly heat. Use the left and right arrow keys with a keyboard.