Google's Gemini Decision 1 answers yes/no questions about video and audio

single source· 1 articles · confidence: medium · first seen 2026-11-11 22:03 UTC

What this means for you

If you already route yes/no questions to Jev or GPT-6 Luna, Gemini Decision 1 is the only one of the three that accepts video and audio directly, at OpenAI's 10 cents per million input tokens. No accuracy comparison has been published, so test it on your own footage before switching.

Google has released Gemini Decision 1, an API model that returns structured answers — yes/no, multiple choice, and rubric scores — instead of writing prose. It is the third entry in that category, after Jev and OpenAI's GPT-6 Luna, and matches OpenAI's price at 10 cents per million input tokens with output free; Jev is cheapest at 4.2 cents per million. Gemini is the only one of the three that takes video and audio as well as text, images and PDFs. Simon Willison's test on a single video returned probabilities rather than the flat 1.0 or 0.0 the other two return.

Models in this story

Key facts

  • ·Gemini Decision 1 supports three question types: yes/no, multiple choice, and rubric scoring. source
  • ·Gemini Decision 1 is priced at 10 cents per million input tokens, with output free — the same as OpenAI's GPT-6 Luna decision model. source
  • ·Jev is the cheapest of the three at 4.2 cents per million input tokens. source
  • ·Jev accepts text and JSON only, OpenAI's model accepts text and images, and Gemini Decision 1 accepts text, images, audio, video and PDFs. source
  • ·Jev and OpenAI return 1.0 or 0.0 for yes/no questions, while Gemini returned probabilities. source
  • ·In Willison's test on one video, Gemini returned probabilities of 0.92 for robot, 0.92 for USA, 0.05 for dragon and 0.08 for France. source

What the sources say

  • Simon Willison — Hands-on test against one robot video, plus a three-way comparison of price, inputs and output format.

Sources

The original reporting. Follow these — they did the work.

← the wire