Stock Markets February 19, 2026

Google Launches Gemini 3.1 Pro, Boosting Reasoning Performance for Complex Tasks

New release significantly improves core reasoning on ARC-AGI-2 and is rolling out in preview across developer, enterprise and consumer channels

By Leila Farooq
Google Launches Gemini 3.1 Pro, Boosting Reasoning Performance for Complex Tasks

Google has introduced Gemini 3.1 Pro, an updated AI model focused on advanced reasoning for science, research and engineering use cases. The model more than doubled reasoning performance over the previous Gemini 3 Pro on the ARC-AGI-2 benchmark, scoring 77.1%. The release is entering a preview phase for developers, enterprises and selected consumer products, with certain features gated to Pro and Ultra plan subscribers.

Key Points

  • Gemini 3.1 Pro achieved a verified ARC-AGI-2 benchmark score of 77.1%, more than doubling the reasoning performance of the previous 3 Pro model.
  • The preview rollout is available to developers via Gemini API in Google AI Studio, Gemini CLI, Google Antigravity and Android Studio; enterprises via Vertex AI and Gemini Enterprise; and consumers via the Gemini app and NotebookLM.
  • Higher usage limits and exclusive NotebookLM access are reserved for Google AI Pro and Ultra plan subscribers, affecting developer, enterprise and consumer-facing AI services.

Google announced Gemini 3.1 Pro, an upgraded iteration of its Gemini 3 series built to tackle sophisticated reasoning tasks commonly encountered in scientific, research and engineering domains.

On the ARC-AGI-2 benchmark, which measures a model's facility for resolving novel logic patterns, Gemini 3.1 Pro recorded a verified score of 77.1%. That result represents more than a twofold increase in reasoning performance compared with the prior 3 Pro release.

The rollout begins with a preview targeted at developers and is available through multiple tooling options. Developers can access the model via the Gemini API in Google AI Studio, the Gemini CLI, the agentic development environment Google Antigravity, and Android Studio. For enterprise customers, Gemini 3.1 Pro is accessible through Vertex AI and Gemini Enterprise. Consumers will be able to use the model through the Gemini app and NotebookLM.

Google highlighted a set of capabilities where the model shows particular strength. Gemini 3.1 Pro is described as adept at tasks requiring deeper reasoning, including producing visual explanations of complex topics, synthesizing data across inputs, and aiding creative workflows. One demonstrated application, labeled "complex system synthesis," involved integrating complex application programming interfaces with clearer design outputs, exemplified by a live aerospace dashboard that visualized the International Space Station's orbit.

Access terms include higher usage limits for Gemini app subscribers on Google AI Pro and Ultra plans. NotebookLM usage of Gemini 3.1 Pro is limited to Pro and Ultra subscribers exclusively.

Google characterized this preview release as a validation stage to confirm the updates prior to a broader general availability release. The company also indicated intentions to continue refining the model, with additional improvements planned for agentic workflows and other areas.


Summary of availability and capabilities

  • Developers - preview access via Gemini API in Google AI Studio, Gemini CLI, Google Antigravity and Android Studio.
  • Enterprises - access through Vertex AI and Gemini Enterprise.
  • Consumers - available in the Gemini app and NotebookLM, with NotebookLM restricted to Pro and Ultra subscribers.

The preview phase and subscription-based access are central elements of this release strategy as Google assesses the update before full deployment.

Risks

  • Preview deployment means the release is subject to validation and further changes before general availability, creating uncertainty for users relying on stable production APIs - impacting enterprise AI and developer tool adoption.
  • Exclusive access to NotebookLM and higher usage limits for Pro and Ultra subscribers could restrict broader consumer adoption and experimentation - affecting consumer AI product uptake and related subscription revenue dynamics.
  • Planned improvements for agentic workflows and other areas indicate the model is still evolving, so current capabilities and integrations may change as Google refines the system - relevant to organizations planning long-term integrations.

More from Stock Markets

OpenAI Narrows Long-Range Compute Plan to $600 Billion, Reframes Growth to Revenue-Linked Spending Feb 20, 2026 Moody's Moves Amazon Outlook to Stable as Company Embarks on Massive AI-Driven Capex Push Feb 20, 2026 Phil Spencer to Retire After 38 Years; Asha Sharma Named CEO of Microsoft Gaming Feb 20, 2026 S&P Moves Amentum to BB, Cites Debt Paydown and Strong Backlog; Outlook Stable Feb 20, 2026 WarrenAI Ranks Five Solar Stocks Poised for Diverse Risk-Reward Profiles in 2026 Feb 20, 2026