Google releases Gemini 3.1 Pro AI model with improved reasoning
Summary
Google releases Gemini 3.1 Pro AI, boasting improved reasoning and problem-solving, with better benchmark scores but still trailing some rivals in user-voted rankings.
Google releases Gemini 3.1 Pro
Google has released Gemini 3.1 Pro, the latest version of its flagship AI model. The update is rolling out in preview today for developers and consumers.
The company says the new model offers better problem-solving and reasoning capabilities. It also served as the "core intelligence" behind recent improvements to Google's Deep Think tool.
Benchmarks show mixed results
Google's announcement includes a slate of new benchmark scores. In the advanced knowledge test Humanity's Last Exam, Gemini 3.1 Pro scored a record 44.4 percent.
That's up from Gemini 3 Pro's 37.5 percent and OpenAI's GPT-5.2 score of 34.5 percent. The new model also showed a major leap in the ARC-AGI-2 logic test.
- Gemini 3 Pro scored 31.1 percent on ARC-AGI-2.
- Gemini 3.1 Pro more than doubled that, reaching 77.1 percent.
However, the model did not immediately top the popular Arena leaderboard. For text, Claude Opus 4.6 currently edges it out by four points.
Focus on graphics and simulations
Google highlighted the model's improved ability to generate graphics and simulations. A comparison video shows more elegant SVG outputs from the new version.
Developers using Gemini for agentic workflows are likely to see improvements. The model nearly doubled its score in the APEX-Agents benchmark.
For regular users asking abstract questions, Gemini 3.1 Pro will probably produce more detailed and nuanced answers than its predecessor.
Availability and pricing
The updated model is available starting today in several platforms. It's coming to AI Studio and the Antigravity IDE in preview.
Enterprise users will find it in Vertex AI and Gemini Enterprise. Regular consumers can access it through the Gemini app and NotebookLM.
The API cost for developers remains unchanged at $2 per 1M input tokens and $12 per 1M output tokens. The context window also stays the same at 1M input and 64k output tokens.
Following Google's typical pattern, a 3.1 update for its faster, cheaper Flash model will likely arrive soon.
Related Articles
India’s top telco tackles AI with $110 billion build plan and proven fast market dominance playbook
Reliance Jio plans to invest $110 billion in AI datacenters over seven years, aiming to make AI services as affordable as it made mobile data in India.
Pi for Excel: AI sidebar add-in for Excel, powered by Pi
Pi for Excel is an open-source AI sidebar for Excel. It reads and edits workbooks using models like GPT or Claude, with tools for formatting, extensions, and recovery.
Stay in the loop
Get the best AI-curated news delivered to your inbox. No spam, unsubscribe anytime.
