New Describe a workflow. Raq.com builds it for you and runs it on autopilot. Try Automations free →
Raq.com
raq.com/tools/ai-pro-reasoning
AI Pro Reasoning interface

Swipe to explore

Access GPT-5.4 Pro reasoning without a subscription. Save $50-$200/month depending on your usage. Ideal for deep research, complex analysis, and detailed reports when you need the most capable model. Attach documents or pull from your Document Library for analysis. Run two models side by side in Dual mode, ask follow-up questions on any report, and share results with a link.

ARC-AGI-2 Scores

Higher is better

arcprize.org

Last checked: 20 Feb 2026

Gemini 3.1 Pro
88.1%
Claude Opus 4.6
79.0%
Claude Sonnet 4.6
65.7%
GPT-5.4 Pro
51.7%
Gemini 3 Flash
34.0%
Opus 4.5 (32K)
28.2%
o3 (High)
2.9%

Gemini 3.1 Pro leads with 88.1% on ARC-AGI-2, the hardest AI reasoning benchmark

The ARC-AGI benchmark tests genuine reasoning and problem-solving, not just pattern matching. It's considered one of the most challenging tests for measuring AI intelligence.

Ready to try AI Pro Reasoning?

Start saving time and money with Raq.com.

Free trial included
No credit card needed
Your prompts stay private