The AI Loadout
←Back to models
Checked Active: 2026-10-19
Perplexity

Perplexity Decider v1.2

Classification, routing, yes/no checks and rubric scoring inside apps and agents. It returns probabilities your code can act on, not generated text.

Evaluation Analysis: Released Oct 16, 2026 on Perplexity's Decisions API (POST /v1/decisions, model pplx-decider-v1.2) and as Apache 2.0 open weights. 27B, built on Qwen3.8-27B, text and image input. One request can ask several questions about the same input (choice, yes/no or score) for about the cost of one. Perplexity reports 68.09 on its Decision Index 0.3, up from 62.30 for v1.1. The 4B pplx-decider-flash-v1.2 costs $0.01 per 1M input tokens and scores 58.88 (60.05 asked one question at a time). Self-hosting the 27B needs a CUDA GPU with room for about 52 GiB of BF16 weights; the 4B needs about 9 GiB. Now also in the Ollama library as pplx-decider: a 17 GB Q4_K_M build with a 256K context that runs through Ollama's /v1/systemone endpoint or ollama run pplx-decider --ask. The MLX builds for Apple Silicon (19 to 52 GB) rely on support added in Ollama 0.40.7, still a pre-release on Oct 19, 2026.
Context Window262K tokens per request (trained on up to about 32K)
Input Price / 1M$0.02
Output Price / 1MFree (no output tokens)
Release Date2026-10-16

Specs and prices checked 2026-10-19 against the official page.

Used in these setups

Related skills and personas

Agent skills and bot personas that pair well with Perplexity Decider v1.2.

Browse full directories:All Skills (>35)·All Personas (>15)

Want tested skills and personas installed in one command? See the AI Loadout packs, from $9.