[KongchangAI]
Modelparallel/parallel-decisions-fast

Parallel Decisions Fast

A real-time web decision model introduced by Parallel that uses retrieval-augmented generation to search the live web. For dynamic factual questions it returns structured answers with probabilities and cited evidence, and it supports boolean, multiple-choice and scoring question types.

Timeline (last 90 days)

Nov 13

Developers can call the Parallel Decisions Fast model for free through AI Gateway, the model name is parallel/parallel-decisions-fast, and no separate Parallel account registration is required.

Unverified50%
Nov 13

Parallel Decisions Fast is free to use on AI Gateway until November 23, 2026.

Unverified50%
Nov 13

Parallel Decisions Fast does not generate conversational text; instead, it searches the web in real time and returns structured answers with probabilities and cited evidence.

Unverified50%
Nov 13

Parallel Decisions Fast supports three structured question types: Boolean (returns the probability of 'yes'), Choice (selects from custom options), and Score (scores according to an ordered rubric).

Unverified50%
Nov 13

Multiple related questions can be posed in a single request, and they share context and retrieved evidence.

Unverified50%
Nov 13

The cited evidence applies to the entire request rather than mapping each source individually to a single answer.

Unverified50%
Nov 13

The AI SDK provides the experimental_decide function for calling Parallel Decisions Fast, and the experimental_ prefix indicates that the API is still experimental and its signature may change.

Unverified50%
Nov 13

Developers should first evaluate the model's accuracy and confidence distribution across question types using a batch of samples with known answers, and then set thresholds.

Unverified50%
Nov 13

When includeExcerpts is enabled, the returned results include evidence excerpts.

Unverified50%

All Facts (9)

Source Articles