---
title: 'Gemini: A Family of Highly Capable Multimodal Models'
license: CC-BY-4.0
license_url: https://creativecommons.org/licenses/by/4.0/
license_note: 'paper is CC BY 4.0 (verified on arXiv abs page 2026-11-12; icon by-4.0.png, http://creativecommons.org/licenses/by/4.0/); note is our own summary'
source: https://iinventedlight.com/library/notes/2312-11805-gemini.md
source_license: 'CC BY 4.0'
attribution: 'Summary by iinventedlight (AI-assisted). Source: Gemini Team and 1350 coauthors (listed on the arXiv abs page), Gemini: A Family of Highly Capable Multimodal Models (CC BY 4.0).'
author: 'iinventedlight (AI-assisted)'
date: 2026-11-12
status: published
approved_by: "Cole Withington, 2026-11-12 14:44 PT, Librarian chat widget t51s316: \"Yes, publish all 100 second-list arXiv notes\""
checked_on: 2026-11-12
as_of: 2026-11-12
confidence: medium
untested: 'Not reproduced. Abs page and HTML key sections read 2026-11-12. Named code URLs checked for HTTP 200; code not run.'
category: arxiv
topic_area: 'ai-research: foundations'
tags: arxiv, foundations, multimodal
arxiv_id: 2312.11805
alias: 2312.11805-gemini
sources:
  - id: S1
    title: 'Gemini: A Family of Highly Capable Multimodal Models'
    publisher: arXiv
    authors: 'Gemini Team and 1350 coauthors (listed on the arXiv abs page)'
    url: https://arxiv.org/abs/2312.11805
    version_read: 'v5 (9 May 2025)'
    accessed: 2026-11-12
    license: 'CC BY 4.0'
---
# Gemini: A Family of Highly Capable Multimodal Models

## Summary
This report introduces Gemini, a family of multimodal models with capabilities across image, audio, video, and text understanding. The family consists of Ultra, Pro, and Nano sizes, intended for uses ranging from complex reasoning to on-device memory-constrained settings. The authors report that Gemini Ultra advances the state of the art in 30 of 32 evaluated benchmarks, including being the first model to achieve human-expert performance on MMLU, and improves the state of the art on every one of the 20 multimodal benchmarks they examined. [S1]

## Claims
- id: C1
  claim: 'The Gemini family consists of Ultra, Pro, and Nano sizes.'
  source: S1
  evidence: "The Gemini family consists of Ultra, Pro, and Nano sizes"
- id: C2
  claim: 'They report that Gemini Ultra advances the state of the art in 30 of 32 benchmarks.'
  source: S1
  evidence: "advances the state of the art in 30 of 32 of these benchmarks"
- id: C3
  claim: 'They report it was the first model to achieve human-expert performance on MMLU.'
  source: S1
  evidence: "the first model to achieve human-expert performance on the well-studied exam benchmark MMLU"

## Verification
- Code available: No official model or training-code URL was named on the abs page or HTML. Demos at https://deepmind.google/gemini returned HTTP 200 and redirected to https://deepmind.google/models/gemini/ on 2026-11-12. Not run.
- Reproduced: not reproduced (2026-11-12). Code not run.
- Related work: Live notes [2403-05530-gemini-1-5](https://iinventedlight.com/library/notes/2403-05530-gemini-1-5/) and [ao-context-million-token-windows](https://iinventedlight.com/library/notes/ao-context-million-token-windows/).
- Limits: Company technical report, not an independent audit. The 30-of-32 and MMLU claims are author-reported. This note follows the v5 text last revised 9 May 2025.
- License: CC BY 4.0 on the abs page (2026-11-12; icon by-4.0.png; http://creativecommons.org/licenses/by/4.0/). Submitted 19 Dec 2023 (v1); last revised 9 May 2025 (v5). Sources: [S1] https://arxiv.org/abs/2312.11805, v5 (9 May 2025). checked_on: 2026-11-12.

## When to apply
Use when tracing the original Gemini family (Ultra, Pro, Nano) and its reported benchmarks.

## Facts
- id: F1
  claim: 'Gemini is introduced as a multimodal family with Ultra, Pro, and Nano sizes.'
  confidence: medium
  date: 2026-11-12
  sources: [S1]
  license: CC BY 4.0
- id: F2
  claim: 'The authors report that Gemini Ultra advances the state of the art in 30 of 32 evaluated benchmarks.'
  confidence: medium
  date: 2026-11-12
  sources: [S1]
  license: CC BY 4.0
- id: F3
  claim: 'They report it was the first model to achieve human-expert performance on the exam benchmark MMLU.'
  confidence: medium
  date: 2026-11-12
  sources: [S1]
  license: CC BY 4.0

## Sources
- [S1] arXiv 2312.11805, https://arxiv.org/abs/2312.11805. License: CC BY 4.0. v5 (9 May 2025). Accessed 2026-11-12.

License: this note is CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Summary by iinventedlight (AI-assisted). Paper is CC BY 4.0.

---

Agents: [/agents.md](/agents.md) | [/.well-known/agent-library.json](/.well-known/agent-library.json).
