Companyfounded in 2022
Perplexity
San Francisco, California, United States · official site ↗
8thout of 10Perplexity, its assistant, Celsius score 36 out of 100
Measured models00 with a profile
Best on intelligencenot measuredno measured model
Figure per questionnothing publishedneither energy nor emissions
Lineage
- Assistant
Perplexity
- Model families
- no measured model
- Models from other companies in Perplexity
OpenAI
Google
Anthropic
SpaceX Super Intelligence
- Headquarters and founding
- San Francisco, California, United States; founded in 2022
- Leadership
- Aravind Srinivas, chief executive and co-founder; Denis Yarats, Johnny Ho, Andy Konwinski, co-founders
Founding and headquarters: Perplexity AI, Form D (SEC, 5/4/2023), 11 November 2026 ↗
Perplexity's assistant
The assistants ranking →Perplexity models
Perplexity does not publish a model measured by Epoch AI or LMArena in 2026. Perplexity serves models from other companies, 10 named on the pages we have read:
- "Best" (automatic model choice)by default, Free and paid plans (official page ↗)
GPT-6.1 SolOpenAI, paid plan, Pro, Max (official page ↗)
GPT-6 AstraOpenAI, paid plan, Max (official page ↗)
Gemini 4 ArgonGoogle, paid plan, Max (official page ↗)
Gemini 3.8 FlashGoogle, paid plan, Pro, Max (official page ↗)
Claude Sonnet 5.5Anthropic, paid plan, Pro, Max (official page ↗)
Claude Opus 5.5Anthropic, paid plan, Max (official page ↗)
Claude Fable 5.1Anthropic, paid plan, Max (official page ↗)
Grok 4.7SpaceX Super Intelligence, paid plan, Pro, Max (official page ↗)
- Kimi K3 (Moonshot AI), GLM 5.3 (Z.ai), Nemotron 3.5 Super VL (NVIDIA)paid plan, Pro, Max (official page ↗)
Governance and safety
- Company governance
- Serves other companies' models; sued by at least 6 media outlets and platforms, accused by Cloudflare of getting around blocks (8/2025).
- Model safety
- No safety framework, no Seoul commitments, no European code; absent from independent evaluations.
Electricity and environmental approach
- Decarbonised electricity
- Rented clouds (AWS, Azure), countries and electricity not published.
- Company's environmental approach
- No figure, no report, no published optimisation.
Methodology: the ranking criteria ; the electricity use of a task for its models is a Projet Celsius estimate, based on the electricity use ofwriting 1,000 tokens (AI Impact Database).
Scores and facts: assistants ranking, scores as of 11/11/2026 (methodology 1.1), sources on the Perplexity page. Models: Epoch AI (CC BY 4.0).
Continue in the ranking
The full ranking →The hub newsletter
This month's ranking, news on the impact of AI and one figure from the AI Impact Database. One newsletter a month at most, free of charge. Past issues








