AI Pro Reasoning: Now Powered by GPT-5.4 Pro
AI Pro Reasoning on Raq.com now uses GPT-5.4 Pro. It builds on the existing workflow for document analysis, follow-up questions, Dual mode, and shareable reports, now with a stronger model behind the work.
Better at real work
OpenAI describes GPT-5.4 as its most capable model for professional work. Improvements span spreadsheets, presentations, documents, and deep web research. That makes this upgrade especially relevant for long reports, technical comparisons, research-heavy tasks, and structured business analysis.
Benchmarks that matter
These are the most relevant public Pro-to-Pro comparisons from OpenAI's GPT-5.4 release materials. All figures are GPT-5.4 Pro vs GPT-5.2 Pro.
| Benchmark | GPT-5.4 Pro | GPT-5.2 Pro |
|---|---|---|
| GDPval (economically useful knowledge work) | 82.0% | 74.1% |
| Investment banking modelling (OpenAI internal) | 83.6% | 71.7% |
| BrowseComp (multi-step web research) | 89.3% | 77.9% |
| GPQA Diamond | 94.4% | 93.2% |
| Humanity's Last Exam (with tools) | 58.7% | 50.0% |
Source: OpenAI GPT-5.4 release materials. BrowseComp was measured at a later date and also reflects changes in the search system.
All five comparisons point the same way: stronger business work, stronger research, and stronger multi-step reasoning.
What you should notice
Expect better synthesis across large sets of information, stronger structure in long reports, and better performance when a task depends on careful research rather than a single quick answer. This is an inference from OpenAI's published evaluations and capability notes, not a direct Raq.com internal measurement.
Same workflow, better output
Raq.com already positions AI Pro Reasoning for deep research, complex analysis, and detailed reports. GPT-5.4 Pro raises the ceiling on that work without changing what the tool is for.
Get started
Start your free trial. No credit card required. Open AI Pro Reasoning and try GPT-5.4 Pro on your next research brief, comparison, or analysis task. Same workflow. Better reasoning.