By Diego Almada Lopez
Publication Date: 2026-11-10 17:47:00
The model posted a 0.857 on Perplexity’s WANDR research benchmark, ahead of GPT-6 Astra’s 0.735
Perplexity has added Google’s Gemini 4 Carbon to its lineup. The model is now available inside Perplexity Computer for both Pro and Max subscribers.
Max subscribers get an extra perk. They can also use Gemini 4 Carbon in Perplexity Search, which turns the top tier into a slightly roomier VIP section.
What subscribers actually get
Inside Perplexity Computer, Gemini 4 Carbon can power two kinds of work. Users can build persistent Teammates or spin up one-off agentic sessions.
The rollout landed on November 10, 2026. Pro users get the model in Computer only, while Max users get it across both Computer and Search.
Gemini 4 Carbon is built for long-horizon workflows. That means tasks with many steps that unfold over time, rather than a quick question and answer. It reportedly shines in cybersecurity and complex reasoning tasks.
The benchmark scorecard
CEO Aravind Srinivas pointed to the model’s showing on Perplexity’s own Wide-And-Deep-Research benchmark, known as WANDR. Gemini 4 Carbon scored 0.857, which the company calls a new state-of-the-art result.
For comparison, GPT-6 Astra scored 0.735 on the same test. That is a meaningful gap on a scale where the ceiling is 1.0.
A caveat worth noting: WANDR is Perplexity’s benchmark. Companies grading on their own curve is not unusual in AI, but it is worth keeping in mind when the scoreboard is…