Cursor Introduces Pro Tier Model Caching and Expanded Context Windows
Cursor upgraded its backend inference engine with smart prompt caching and expanded 128k context support on Claude 3.5 Sonnet.
Verified State Diff
Comparison Mode:
- Previous State
3–5 second latency on large codebase context indexing with standard token pricing.
+ Verified New State
Sub-second warm completions with prompt caching enabled across Pro requests.
Impact & Verification Analysis
WHO IS AFFECTED
All Cursor Pro and Enterprise developers working on large repositories.
WHY IT MATTERS
Eliminates developer flow interruption during complex refactorings.
Full Fact Overview
Fast requests on the Pro plan now execute with sub-second time-to-first-token by caching project repository structural trees.
TRACKED ENTITY
Explore all historical Cursor changes