Google on Thursday announced Gemini 3.1 Pro, a new large language model described as a step forward in core reasoning, and published benchmark results showing a 77.1% score on ARC-AGI-2 versus 31.1% for Gemini 3 Pro. Google said Gemini 3.1 Pro outperforms several commercial rivals on many cited tests, though some benchmarks favor Anthropic or OpenAI on specific suites. The model is available via Gemini API, Google AI Studio, Vertex AI, and integrated Azure developer tools.
Key Points
- 1Gemini 3.1 Pro scores 77.1% on ARC-AGI-2, outperforming prior Gemini variants.
- 2Highlights marked reasoning improvement across multiple benchmarks versus Opus and GPT-5 variants.
- 3Enables developers and enterprises to access via API, Vertex, and Microsoft tools for integration.
Scoring Rationale
Significant model improvements and broad availability boost impact, but reliance on vendor-reported benchmarks limits independent confidence.
Sources
Public references used for this report.
Practice with real Ad Tech data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Ad Tech problems
