Perplexity says it has released pplx-decider-v1.1-27b, an updated open-weights multimodal decision model. In the same announcement, the company said the model “scores the highest” on Hugging Face’s Decision Index 0.3 benchmark and that its input pricing is half that of v1, at $0.02 per million input tokens. Perplexity CEO Arav Srinivas separately repeated those points in posts about the benchmark result and the pricing cut (benchmark claim, pricing claim).
Taken together, that is Perplexity’s pitch for the update: an open-weights model built for decision tasks that the company says improves its standing on a relevant benchmark while cutting usage costs.
OpenRouter added another way to access it
A few days later, OpenRouter said Perplexity Decider v1.1 was available on its platform. In that post, OpenRouter said the model could take text, JSON, or image inputs, return typed answers with probabilities, cost $0.02 per million input tokens, and offer free output. Perplexity’s developer account then pointed users to the OpenRouter listing.
That update matters mainly as distribution news: beyond the original release, OpenRouter presented the model as another option for developers who want to try it through its service.
The later clinical claims are still just that here
The supplied packet also includes later social posts making more specific claims about clinical-decision performance. One post said pplx-decider v1.1 got 643 of 669 right versus 628 for Jev, at 42% lower cost per decision and about the same speed (Maziyar Panahi’s post). Srinivas later amplified that line with a post calling Perplexity’s model the “best quality and lowest cost for clinical decisions” (here).
In the supplied source packet, though, those remain attributed social posts rather than independently verified results. The clearest supported development here is the model release itself, Perplexity’s benchmark and pricing claims, and the later OpenRouter availability.
As presented in these posts, Perplexity is positioning pplx-decider-v1.1-27b as a cheaper follow-up to v1 that it says performs well on decision benchmarks, with broader access through OpenRouter following shortly after release.
Comments