Perplexity launches a Decisions API and open-sources pplx-decider-v1-27b, a decision model it says edges Jev (85.71% vs 84.51%)
Confirmed
The takeaway
On Oct 1, 2026, hours after Cloudflare’s Clef, Perplexity released the Decisions API and the Apache-2.0 weights of pplx-decider-v1-27b, a multimodal model that returns a probability distribution over a fixed set of answers instead of text.
Status
- Claim
Confirmed
- Our reporting
- High confidence
- Importance
- 2 of 5
- Last verified
- 10 October 2026
Your AI and this story
- GPT-6.2 Astra123 days after its cutoff
- Claude Opus 5.593 days after its cutoff
- Gemini 4 Hydrogen184 days after its cutoff
- Grok 4.7123 days after its cutoff
None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 93 days before it.
Key facts
- Model: pplx-decider-v1-27b, fine-tuned from Qwen3.8-27B; text and image input; Apache 2.0 weights on Hugging Face
- API: POST https://api.perplexity.ai/v1/decisions; $0.04 per 1M input tokens, output tokens free; up to 262,144 input tokens and 1–128 questions per request
- Perplexity’s panel (11 benchmarks): overall 85.71% vs Jev 84.51% and base Qwen3.8-27B 74.76%; biggest gain on RAGTruth (88.80% vs 77.27%); Jev still leads on WinoGrande (90.70% vs 83.30%), BBH (94.27% vs 82.80%), JudgeBench, JevBench public hard and others
- Launch post by @perplexitydevs ~242k views on X
- The New Stack (Oct 9): TypeSafe’s ‘System One’ schema is becoming the interoperability contract for decision models: AWS’s open-weight Strands Decider, Upstage’s Solar Decide and Ollama (from v0.35) implement the same /v1/systemone endpoint; OpenRouter, Venice and Perplexity keep the semantics on their own endpoints; OpenAI is the holdout with its own closed Decisions API (GPT-6 Luna). Three question types: choice, score and yes/no. (Article body could not be fetched; facts from the search-result summary of the article, confidence medium)
What happened
Perplexity followed TypeSafe’s Jev (Sept 15), OpenAI’s announced Decisions API and Cloudflare’s Clef with its own decision model: a single forward pass scores every allowed option for typed questions (choice, score), so outputs are calibrated probabilities rather than generated text. The weights are open.
Why it matters
Within three weeks of Jev’s launch, at least four large companies shipped compatible or competing “System One” decision models, two of them open-weight, which turns a single startup’s product into a model category. All benchmark numbers are Perplexity’s own.
Sources
5 sources from 5 sites. Numbers match the chips in the text.
5 sources: 3 primary, 2 press
Primary
- Perplexity docs: Decisions API quickstartdocs.perplexity.ai, docs
- Hugging Face: perplexity-ai/pplx-decider-v1-27bhuggingface.co, code
- Perplexity Developers on X: Introducing the Decisions APIx.com, official
Press
Changes
- Added The New Stack’s Oct 9 overview of System One API convergence (body not fetchable, from search summary)
- Filed (AI Weekly)