Google releases Gemini 1.5 Pro with a million-token context window

事件摘要

On February 15, 2024, Google released Gemini 1.5 Pro, the first large language model with a 1 million token context window — later expanded to 2 million. This breakthrough allowed the model to process entire codebases, hundred-page documents, and hours of video in a single prompt, achieving a 'needle in a haystack' retrieval accuracy of 99.7% across the full context. It set a new capability frontier that competitors raced to match throughout 2024.

影响评估

• Capability Leap +2 · Short-term

  10x+ context window expansion from standard 128K to 1M tokens. 99.7% needle-in-a-haystack accuracy. Enabled entirely new use cases: full-codebase analysis, hundred-page document reasoning, long-form video understanding in a single prompt.

  Affected Groups: developers, researchers, enterprise users, legal professionals

• Paradigm Shift +1 · Medium-term

  Triggered industry-wide race to longer context windows. By end of 2024, every major AI lab was competing on context length as a key capability metric alongside reasoning and multimodality.

  Affected Groups: AI companies, researchers

共识度与来源

• 1
  Introducing Gemini 1.5, Google's next-generation AI model — Google Blog (February 15, 2024)

  URL: https://blog.google/innovation-and-ai/products/google-gemini-next-generation-model-february-2024/

  Gemini 1.5 Pro delivers dramatically enhanced performance, with a breakthrough in context-window length — up to 1 million tokens consistently.

  Reference Evidence Citation logged Live source