Google announces "Gemini 1.5" — 1M-token context handles full-length video and large codebases at once
Google has announced its next-generation model, “Gemini 1.5.” The headline feature is a dramatically expanded context window, reaching up to 1 million tokens — an extraordinary length for its time. The model also drew attention for matching the performance of Gemini 1.0 Ultra while using significantly fewer compute resources.
Details
- Context window: Supports up to 1 million tokens by default — enough, Google said, to process roughly 700,000 words of text or about an hour of video in a single pass
- First model released: “Gemini 1.5 Pro” launched in private preview for developers and enterprises via Google AI Studio and Vertex AI
- Architecture: Adopts a Mixture-of-Experts (MoE) design, improving efficiency while matching the performance of 1.0 Ultra
- Example use cases: Promoted for previously difficult tasks such as understanding the content of long-form video, analyzing entire large codebases, and searching across lengthy documents
- Availability: Launched with limited early access rather than general availability
What happened next
Gemini 1.5 was expanded that May with the addition of a lighter “1.5 Flash” variant, rounding out a lineup balancing long context with speed. Long-context processing went on to become a defining differentiator for the Gemini series, with the performance race continuing through “Gemini 2.0” in December 2024 and “Gemini 2.5” in March 2025.