Perplexity just gave free users a 1M token context window

If you've ever tried feeding a 300-page PDF to an AI only to get an error saying your file is too long, Perplexity just fixed that headache for you without asking for a credit card.

In a LinkedIn post on Friday, the AI search company announced that it has expanded the context window for file and image uploads to 1 million tokens for all free users with an account.

To put 1 million tokens into perspective, that is roughly enough room to process an entire 350-to-400-page book, around 1,000 pages of monthly financial PDFs, or between 25,000 and 50,000 lines of code in a single go.

For comparison, OpenAI caps ChatGPT's GPT-4o model at 128,000 tokens, and that is a tier reserved for paying subscribers.

This move comes during a fierce race among AI search providers.

ChatGPT recently rolled out its search functionality and its latest o3-mini model for free to stay competitive against rivals like DeepSeek, alongside tools like Deep Research.

While ChatGPT aims to be an all-in-one assistant for coding, custom bots, and general conversation, Perplexity stays laser-focused on source-backed research and information retrieval.

Why it matters

Offering a 1 million token context window at zero cost is a massive win for students, academics, and researchers who regularly drown in dense reports.

Instead of chopping up massive whitepapers or worrying about context truncation, anyone with a free login can now throw an entire manual or quarterly earnings report into Perplexity and query it directly.

We think this is a brilliant counter-attack in the search wars. By offering eight times the context capacity of ChatGPT's paid tier for free, Perplexity puts immediate pressure on OpenAI and Google to reconsider how they gate large-context features.

The real question is sustainability. Ingesting 1 million tokens per upload is computationally heavy, and running that for a massive free user base could create serious server bills.

Still, if you need an alternative search engine that can actually read your longest documents without charging twenty bucks a month, it is well worth testing right now.