Gemini 3.1 Pro set a new standard on ARC-AGI 2, the benchmark designed to measure genuine reasoning ability. It is available as a model option in AI Pro Reasoning on Raq.com.
ARC-AGI 2 is a benchmark built to test whether AI models can reason through novel problems they have never seen before. Unlike benchmarks that reward pattern matching or memorisation, ARC-AGI 2 presents tasks that require flexible, general-purpose thinking. Each problem is unique, so a model cannot rely on having seen similar examples during training.
Scoring well on ARC-AGI 2 signals that a model can adapt to unfamiliar situations, break down complex problems, and arrive at correct answers without shortcuts. It is one of the clearest measures of raw reasoning capability available today.
Gemini 3.1 Pro achieved a strong result on ARC-AGI 2, outperforming previous models on tasks that require multi-step logical reasoning and abstraction. This is not incremental improvement. It represents a meaningful step forward in how well a model handles problems it was not explicitly trained to solve.
For users working with AI Pro Reasoning, this translates to better handling of complex analysis, research questions that require genuine inference, and tasks where the answer is not a straightforward lookup.
AI Pro Reasoning on Raq.com runs a multi-stage pipeline to produce thorough, well-researched outputs. It routes your question through an analysis stage, optionally searches the web for supporting information, then passes everything to a reasoning model for a detailed response. You can also enable an engaging summary, fact-checking, and AI-generated visuals.
You can choose from two reasoning models:
Select your preferred model from the dropdown before submitting a query. The cost estimate updates in real time so you know what to expect.
Gemini 3.1 Pro is a strong choice when your query requires genuine reasoning rather than retrieval. If you are asking the model to work through a problem step by step, compare multiple options with trade-offs, or analyse something it has not seen before, Gemini 3.1 Pro's ARC-AGI 2 results suggest it will handle that well.
For straightforward research and long-form writing, GPT-5.4 Pro remains a reliable default. Having both in one tool means you can pick the right model for each task.
AI Pro Reasoning is included in your Raq.com subscription. Gemini 3.1 Pro is competitively priced alongside GPT-5.4 Pro. You see the estimated cost before you submit.
Start your free trial . No credit card required. Open AI Pro Reasoning from your dashboard, select Gemini 3.1 Pro from the model dropdown, and put it to work on your next complex question.