context-windows-expanded-10000x-in-seven-years

IN derived (depth 1)

Created 2026-06-21T09:54:53+00:00 · Reviewed 2026-06-21T14:41:08+00:00

Context windows expanded approximately 10,000x from GPT-2's 1,024 tokens (2019) to Llama 4 Scout's 10 million tokens (2025), with Claude 2.1's 200K window as a major intermediate milestone in late 2023.

Summary

AI models' working memory has grown from about a page and a half of text to roughly 30,000 pages in just six or seven years, meaning a single model can now hold an entire book, a full codebase, or months of conversation in active consideration at once. For any system that reasons over information, this collapse of the context bottleneck changes what is practically possible: the constraint of "the model will forget what you said twenty pages ago" is no longer the design driver it once was.

Justifications

SL — Three context window measurements across model families chart an exponential growth trajectory

Antecedents (all must be IN):

Dependents

These beliefs depend on this one: