Subvolts take Google DeepMind details scaling AI context windows to millions of tokens and their impact on agents. May 2, 2025 - Developers
- Learn about scaling context windows, RAG integration, and tips for using long context models effectively.
- Business
- Understand how larger context windows enable more capable AI agents and improve information processing for enterprise applications.
This interview provides a detailed look at significant advancements in AI model capabilities directly from a leading researcher.
Deep Dive into Long Context
Audio streams from Google AI. Your place is remembered on this device.
From Subvolts
Summary and analysis by Subvolts editorial, AI-assisted.
Google DeepMind's Nikolay Savinov joins Logan Kilpatrick on Release Notes to discuss scaling AI context windows to millions of tokens. They explore improvements since the 1.5 Pro release, the role of long context in AI agents, and compare it with Retrieval Augmented Generation (RAG). The conversation also touches on future advancements like near-perfect recall and cost reduction.
This episode explains how AI models are getting much better at remembering and processing long pieces of text, like entire books. This improved memory helps AI agents perform more complex tasks and understand information more deeply.
This episode offers insights into a critical area of AI development: extending the memory of language models. Listeners learn about the technical challenges and potential of very long context windows, and how they enable more capable AI agents. It's for researchers and developers interested in the practical application and future direction of large language models. This directly relates to ongoing efforts by Google, OpenAI, and Anthropic to build more powerful and context-aware AI systems.
- HostLogan Kilpatrick
- GuestNikolay SavinovGoogle DeepMind
- Guest: Nikolay Savinov
- Host: Logan Kilpatrick
- Show: Google AI: Release Notes
- Length: 60 minutes
- Context Window Scale: Millions of tokens
- Product Mentioned: Gemini 1.5 Pro
Explore the synergy between long context models and Retrieval Augmented Generation (RAG) in this episode of Release Notes. Join Google DeepMind's Nikolay Savinov as he discusses the importance of large context windows, how they enable Al agents, and what's next in the field.
Chapters:
0:52 Introduction & defining tokens
5:27 Context window importance
9:53 RAG vs. Long Context
14:19 Scaling beyond 2 million tokens
18:41 Long context improvements since 1.5 Pro release
23:26 Difficulty of attending to the whole context
28:37 Evaluating long context: beyond needle-in-a-haystack
33:41 Integrating long context research
34:57 Reasoning and long outputs
40:54 Tips for using long context
48:51 The future of long context: near-perfect recall and cost reduction
54:42 The role of infrastructure
56:15 Long-context and agents
- tokens
- Nikolay Savinov
- pieces of words or characters that AI models process
- RAG
- Nikolay Savinov
- Retrieval Augmented Generation, a technique to improve AI responses using external data
- context window
- Nikolay Savinov
- the amount of text an AI model can consider at once
27 minKoray Kavukcuoglu on frontier models, coding agents, and building AGI
WithKoray Kavukcuoglu
39 minIntroducing Gemini Robotics 2
WithCarolina Parada, Stuart Bowers, Kanishka Rao, Jie Tan
42 minGemini co-leads on project origins and what's next
WithJeff Dean, Koray Kavukcuoglu, Noam Shazeer, Oriol Vinyals
40 min
37 minAI and the Future of Health with Joelle Barral
WithJoelle Barral
Questions people ask
- What is the difference between RAG and long context?
- The episode explores the synergy between long context models and Retrieval Augmented Generation (RAG), discussing their respective roles and how they compare.
- What are the future goals for long context models?
- Future goals include achieving near-perfect recall and significant cost reduction for processing long contexts.
Page generated Oct 19, 2026. Summaries are Subvolts' own; the audio, artwork and notes belong to Google AI.