← Back to the A-Z glossary
Model · Google · 2026

Gemini 3.1 Pro

Gemini 3.1 Pro is Google DeepMind's big model, released in early 2026, and it powers Gemini Advanced, NotebookLM and Google AI Studio. Its strength: a one-million-token context window, so you can hand it hundreds of pages or hours of video in one go — text, image, audio and video all mixed together.

Model 5 min read Updated 2026-05-25
— What it is

Gemini 3.1 Pro, in plain words

In plain words: it's Google's model built to swallow huge volumes. Where others choke, it digests whole case files in a single request.

Its specialty: a one-million-token context window as standard, stretchable to 2M for heavy use. In practice, I can hand it 10 hours of podcast, 500 pages of PDF, or 50,000 lines of code to analyze in a single request.

It's natively multimodal: it reads text, image, audio, video and code in the same prompt. No need to convert anything first.

What I use it for

  • Analyzing a long video: I give it the URL of a 90-minute talk on YouTube, and it hands me a chaptered summary with timecodes.
  • Reading a full case file: 30 admin documents, invoices, contracts — I drop the lot in, it cross-checks and flags the inconsistencies.
  • NotebookLM in podcast mode: I give it my sources, and NotebookLM generates a 15-minute podcast of two voices discussing them. Mind-blowing.
  • Coding with a big context: I can throw in an entire codebase (up to ~50k lines) and ask it to suggest a full refactor.
  • A draft in French: not the best on tone, but fine when price comes first.

How it compares

vs GPT-5.5: Gemini has a context window 4× bigger and costs 2× less. GPT-5.5 is still better at structured reasoning and conversational tone. Pick Gemini when the volume of data comes first, GPT-5.5 when finesse comes first.

vs Claude Sonnet 4.6: Sonnet writes better and reasons better on nuanced problems. Gemini wins on the giant context and the price. It's a tie on multimodal (both read images and PDFs).

vs Gemini 2.5 Flash: Flash is 10× cheaper but less precise. Pro for tasks that need quality, Flash for volume.

What it costs

API pricing as of May 25, 2026 (source: ai.google.dev/pricing):

  • Input: $2–4 / million tokens (depending on how much context you use)
  • Output: $12–18 / million tokens
  • Higher rate beyond 200k tokens used in input

Through Gemini Advanced (the consumer app): $20/month in the Google AI Premium plan, which also includes 2 TB of Drive.

My take

I use it for two things: NotebookLM (the generated podcasts are incredible for revising a topic) and analyzing big volumes that Claude would refuse.

What I love: the price, the giant context, the quality of the multimodal. What bugs me: the consumer Gemini interface is less ergonomic than ChatGPT or Claude, and the tone is too "Google" — neutral, cautious, no personality.

Not my daily model, but essential for the tough spots where you have to digest a mountain of material.

Quick questions

How many tokens in its context window?

Gemini 3.1 Pro is Google DeepMind's big multimodal model (2026), with a one-million-token context window. It reads text, image, audio and video, and powers Gemini Advanced and NotebookLM.

Does Gemini read YouTube videos?

Yes, by dropping the URL straight into Gemini Advanced or AI Studio.

Gemini or ChatGPT for everyday use?

ChatGPT for conversation and writing, Gemini for chewing through volume.

Verified 2026-05-25 · next review 2026-11-25

Prices and specs verified on Google AI pricing. Personal use: NotebookLM 1–2 times a week, Gemini API for analyzing big corpora.

Want to keep going in order?

Head back to the full glossary, search a word, then open only the pages that deserve more than a short definition.

Open the AI glossary