Should Cursor's Pro plan include a larger premium-model credit pool or more generous Auto mode usage?

Introduction

Cursor Pro costs $20 a month. It includes unlimited Tab, a usage pool for premium models billed at roughly API rates, and Auto mode, which picks a model for the user and is far cheaper for Cursor to serve, especially when it routes to Cursor's own Composer models. Pro+ ($60) and Ultra ($200) mainly offer larger usage pools. Since the July 2025 pricing backlash, every change to what Pro includes is watched closely by the developer community.

The trade-off: with a fixed budget of model cost per Pro user, should Cursor put more into a larger premium-model credit pool (choice of frontier models from OpenAI, Anthropic and Google) or into more generous Auto mode usage?

Step 1

Clarifying Questions (3 minutes)

• Context: is the total cost budget per Pro user fixed? I'll assume yes: Cursor wants Pro's gross margin to stay roughly constant.

• Business: is the goal retention, upgrades to Pro+ and Ultra, or new subscriber conversion? I'll assume retention and healthy upgrades are equally important.

• User impact: how do Pro users split today between Auto and manual premium model selection? I'll assume about 55% of agent requests go through Auto, but the heaviest 20% of users pick premium models manually most of the time.

• Technical: has Auto quality improved with Composer 2 and better routing? I'll assume yes, with keep rates close to premium models on routine tasks but lower on hard ones.

• Resource: can we vary allowances by user cohort in an experiment? I'll assume we can, with clear communication and no one getting less than today.