Before it
Long context on the mid tier was capped at 200K tokens.
The problem
Agents working across large repositories exhaust that quickly.
What it introduced
A mid-tier model with a million-token context in beta.
Why it matters
Long context reached the price tier that agents actually run on.
What modern models inherit
Very long context at mid-tier prices.
What came next
The Claude 5 generation mid-year.