Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context
Summary
This report introduces the Gemini 1.5 family, including an updated Gemini 1.5 Pro and a lighter Gemini 1.5 Flash. Gemini 1.5 Pro is a sparse mixture-of-expert (MoE) Transformer-based model. The report finds near-perfect retrieval (>99%) up to at least 10M tokens, including long documents and hours of video and audio. [S1]
Claims
- id: C1 claim: 'The family includes an updated Gemini 1.5 Pro and Gemini 1.5 Flash.' source: S1 evidence: "an updated Gemini 1.5 Pro, which exceeds the February version"
- id: C2 claim: 'They report near-perfect retrieval above 99% up to at least 10M tokens.' source: S1 evidence: "near-perfect retrieval (>99%) up to at least 10M tokens"
- id: C3 claim: 'They describe Gemini 1.5 Pro as a sparse mixture-of-expert Transformer.' source: S1 evidence: "Gemini 1.5 Pro is a sparse mixture-of-expert (MoE) Transformer-based model"
Verification
- Code available: No official model or training-code URL was named as a release of this paper on the abs page or HTML. Cited third-party repos were not treated as this paper's code.
- Reproduced: not reproduced (2026-11-11). Code not run.
- Related work: Live notes ao-context-million-token-windows and 2404-06654-ruler. This is the paper-verification note for the Gemini 1.5 report itself.
- Limits: The long-context retrieval figures are reported by the authors on their own evaluations. The report is a company technical paper, not an independent audit, and the models themselves were not released as open weights in this paper.
- License: CC BY 4.0 on the abs page (2026-11-11; icon by-4.0.png; http://creativecommons.org/licenses/by/4.0/). Submitted 8 Mar 2024 (v1); last revised 16 Dec 2024 (v5). Sources: [S1] https://arxiv.org/abs/2403.05530, v5 (16 Dec 2024). checked_on: 2026-11-11.
When to apply
Use when asking what Gemini 1.5 reported for long multimodal context, not as a substitute for an independent eval.
Facts
- id: F1 claim: 'The Gemini 1.5 family includes an updated Gemini 1.5 Pro and Gemini 1.5 Flash.' confidence: medium date: 2026-11-11 sources: [S1] license: CC BY 4.0
- id: F2 claim: 'The authors report near-perfect retrieval above 99% up to at least 10M tokens.' confidence: medium date: 2026-11-11 sources: [S1] license: CC BY 4.0
- id: F3 claim: 'They describe Gemini 1.5 Pro as a sparse mixture-of-expert Transformer-based model.' confidence: medium date: 2026-11-11 sources: [S1] license: CC BY 4.0
Sources
- [S1] arXiv 2403.05530, https://arxiv.org/abs/2403.05530. License: CC BY 4.0. v5 (16 Dec 2024). Accessed 2026-11-11.
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.
License: CC-BY-4.0 · Summary by iinventedlight (AI-assisted). Source: Gemini Team, Petko Georgiev, Ving Ian Lei, Ryan Burnell, Libin Bai, Anmol Gulati, et al. (1137 authors), Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context (CC BY 4.0).
Cite this
'iinventedlight (AI-assisted)'. 2026. "Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context." iinventedlight Library, iil:note:zrx2ghsx, version of 2026-11-11. https://iinventedlight.com/library/notes/2403-05530-gemini-1-5/. License: CC-BY-4.0.
@misc{iil_note_zrx2ghsx,
author = {{'iinventedlight (AI-assisted)'}},
title = {{Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context}},
howpublished = {iinventedlight Library},
year = {2026},
date = {2026-11-11},
url = {https://iinventedlight.com/library/notes/2403-05530-gemini-1-5/},
note = {Permanent ID iil:note:zrx2ghsx. License: CC-BY-4.0.},
}Spotted a mistake? Use the MCP feedback tool or email hello@iinventedlight.com.
For language models and agents: start at /llms.txt | /agents.md | /.well-known/agent-library.json. This page in Markdown: /library/notes/2403-05530-gemini-1-5.md.