List your company

Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context

About

In this report, we introduce the Gemini 1.5 family of models, representing the next generation of highly compute-efficient multimodal models capable of recalling and reasoning over fine-grained information from millions of tokens of context, including multiple long documents and hours of video and audio. The family includes two new models: (1) an updated Gemini 1.5 Pro, which exceeds the February version on the great majority of capabilities and benchmarks; (2) Gemini 1.5 Flash, a more lightweight variant designed for efficiency with minimal regression in quality. Gemini 1.5 models achieve near-perfect recall on long-context retrieval tasks across modalities, improve the state-of-the-art in long-document QA, long-video QA and long-context ASR, and match or surpass Gemini 1.0 Ultra's state-of-the-art performance across a broad set of benchmarks. Studying the limits of Gemini 1.5's long-context ability, we find continued improvement in next-token prediction and near-perfect retrieval (>99%) up to at least 10M tokens, a generational leap over existing models such as Claude 3.0 (200k) and GPT-4 Turbo (128k). Finally, we highlight real-world use cases, such as Gemini 1.5 collaborating with professionals on completing their tasks achieving 26 to 75% time savings across 10 different job categories, as well as surprising new capabilities of large language models at the frontier; when given a grammar manual for Kalamang, a language with fewer than 200 speakers worldwide, the model learns to translate English to Kalamang at a similar level to a person who learned from the same content.

Key insights

What problem does Gemini 1.5 address?

The paper studies multimodal models that can recall and reason over fine-grained information across millions of tokens of mixed text, video, and audio. It introduces Gemini 1.5 Pro and a lighter Flash variant, both built to handle up to 10M-token contexts.

How does it differ from prior work?

Gemini 1.5 Pro achieves near-perfect needle-in-a-haystack recall (>99.7%) up to 10M tokens, far exceeding GPT-4 Turbo (128k) and Claude 3 (200k). It also wins 35/45 benchmarks against Gemini 1.0 Ultra while using less training compute.

What are the practical findings useful for?

The paper documents benchmark and prototype capabilities: long-document QA, long-video QA, in-context language learning from a grammar book, and ASR. It also reports estimated productivity time savings across 10 job categories but does not document production deployments.

What limitations and future work remain?

The authors note regressions in audio capabilities for non-head languages, susceptibility to handcrafted prompt injection attacks, and that long-context safety risks may increase as capabilities improve. They call for new benchmarks requiring complex reasoning over long inputs.

Related benchmarks