Writing

Where ChatGPT Pro reasoning fits in a research workflow

• AI Research
• Large Language Models
• Research Workflows

Many regular ChatGPT users may not know that Pro reasoning is available directly in the standard chat screen. It runs GPT-5.6 Sol Pro and is intended for difficult, longer-running work.

Access to Pro in standard chat is included with ChatGPT Pro, Business, and Enterprise plans. Plus users can choose Medium or High reasoning, but not Extra High or Pro. For individual users, ChatGPT Pro starts at $100 per month, and the $200 tier offers higher usage limits. In Business and Enterprise workspaces with flexible pricing, one Sol Pro message uses 50 credits; a Sol message at Medium, High, or Extra High uses 10.

Cover image from the original LinkedIn post.

I find Pro most useful for problems where a superficial answer is genuinely costly: auditing a mathematical formulation, working through a difficult derivation, comparing algorithmic alternatives, or reviewing a codebase against the literature.

The workflow I currently prefer is roughly:

• Sol Pro for deep analysis
• Fable for critique and verification
• Opus or Terra for implementation

The important part is the second step. I would not treat a long analysis from Pro as a decision on its own. I want another model to challenge the assumptions, check the evidence, and look for alternatives before I turn the analysis into code or experiments.

For research, this seems more useful than simply using the most expensive model for every task.