PerplexityDecision modelOpen weights

pplx-decider-flash-v1.2

Summary

pplx-decider-flash-v1.2 is Perplexity's smaller version of its v1.2 decision model. The October 20, 2026 announcement names the same joint request format for multiple questions. A local Mac app for the Flash variant followed on October 23.

Flash is based on Qwen3.5-4B and, according to its model card, was distilled from Perplexity's 27B decision models. Inputs include text and images, which are resized to at most 1.6 megapixels by default. This limit is configurable. Responses contain selections, scores, or probabilities.

Perplexity names its API Console and OpenRouter as access routes, and its local Mac app also uses this model. The model card lists Apache-2.0, and Hugging Face currently lists the checkpoint as public and ungated. Its usage text still contains a conflicting statement that the repository is private.

All benchmark results

Every published result with its source, test setup and comparison models.

Show all 3 published benchmark results

Measurements without matching peer values

These measurements have no matching peer values under the same test conditions. Their original values and sources remain available here.

Decision Index 0.3 (Public, joint requests) (Show measurement, test conditions, and source)

Source value: 58.88
Score: 58.88
Metric: index score
Unit: points
Benchmark version: 0.3
Category: decision
Direction: Higher is better
Source type: vendor-reported
Evaluator: Perplexity pplx-decider-flash-v1.2 model card
Status: active
Retrieved at: 2026-11-07
Methodology: Vendor-reported result. Public tests. Questions in the same request share one sequence within the length limit.

Source Perplexity pplx-decider-flash-v1.2 model card

Decision Index 0.3 (Public, one question per request) (Show measurement, test conditions, and source)

Source value: 60.05
Score: 60.05
Metric: index score
Unit: points
Benchmark version: 0.3
Category: decision
Direction: Higher is better
Source type: vendor-reported
Evaluator: Perplexity pplx-decider-flash-v1.2 model card
Status: active
Retrieved at: 2026-11-07
Methodology: Vendor-reported result. Public tests. Each question is sent separately with its state.

Source Perplexity pplx-decider-flash-v1.2 model card

Vision Decision Index 0.3 (Public) (Show measurement, test conditions, and source)

Source value: 72.53
Score: 72.53
Metric: index score
Unit: points
Benchmark version: 0.3
Category: decision
Direction: Higher is better
Source type: vendor-reported
Evaluator: Perplexity pplx-decider-flash-v1.2 model card
Status: active
Retrieved at: 2026-11-07
Methodology: Vendor-reported result. Public vision tests. The evaluation used images up to 1.6 megapixels.

Source Perplexity pplx-decider-flash-v1.2 model card

Similar models

More models of the same kind as pplx-decider-flash-v1.2, chosen by model line, developer, size, price and release date.

All specifications

Architecture

Model class: Compact decision model with a joint request format for multiple questions.; Source: Model class
Base model: Qwen3.5-4B, distilled from Perplexity's 27B decision models.; Source: Base model
Checkpoint parameters: 4,539,265,536 according to Safetensors metadata.; Source: Checkpoint parameters
Checkpoint: The Hugging Face checkpoint is listed as public and ungated.; Source: Checkpoint

Input and output

Modalities: Text and images as input. Images default to a 1.6 MP size limit that can be configured. Outputs are structured answers, not free-form text generation.; Source: Modalities

Watermarking and provenance

AI watermarking refers to invisible signals embedded in the output. Visible marks and provenance metadata are listed separately. Unknown means there is insufficient evidence for that particular field.

AI watermarking: Unknown; no source
Watermarking technique: Not known; no source

Licence and access

API access: Perplexity API Console and OpenRouter.; Source: API access
License: Apache-2.0 according to the model card.; Source: License
Shutdown: None announced; no source

Further details

Mac app: On October 23, 2026, Perplexity announced an app that runs Flash v1.2 locally.; Source: Mac app
Length limit: The model card gives 262,144 tokens as the maximum sequence length. The vendor says training reached about 32,000 tokens.; Source: Length limit

Prices

Input price: $0.01 per 1M input tokensSource
Output price: Output tokens are free.Source
Web search and URL fetching: Perplexity charges separately for searches and successful URL fetches in web grounding mode.Source

Sources and data date

Benchmark sources at a glance

Which independent evaluators tested this model. This is coverage, not a quality ranking.

No active result from the reviewed external evaluators can currently be attributed to this model. Other source evidence remains available below.

Without vendor claims and historical or disputed results; those stay labelled in the single values.

Type	Evidence and data date
Research status	Research date November 7, 2026. 1 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication.
Additional source	Perplexity pplx-decider-flash-v1.2 model card (retrieved November 7, 2026) · Perplexity pplx-decider-flash-v1.2 model card · Editorial description reviewed November 7, 2026
Additional source	Perplexity v1.2 announcement (retrieved November 7, 2026) · Perplexity v1.2 announcement · Editorial description reviewed November 7, 2026
Additional source	Perplexity local Mac announcement (retrieved November 7, 2026) · Perplexity local Mac announcement · Editorial description reviewed November 7, 2026
Additional source	Perplexity pplx-decider-flash-v1.2 weight metadata (retrieved November 7, 2026)
Additional source	Perplexity Decisions API pricing (retrieved November 7, 2026)
Additional source	Perplexity pplx-decider-flash-v1.2 model card (retrieved November 7, 2026)