Google launches Gemini 3.1 Pro: Reasoning performance doubles in major AI leap | VIDEO

Google on Thursday unveiled Gemini 3.1 Pro, a major update to its artificial intelligence lineup that the company says more than doubles the reasoning performance of its predecessor.

The new model, released in preview, is designed to move AI beyond simple text generation toward "agentic" workflows, autonomous systems capable of solving complex, multi-step problems in fields like software engineering, finance, and research.

The "Step Forward" in Reasoning

The launch of Gemini 3.1 Pro marks a major leap in core intelligence, building on the architecture of the Gemini 3 series released last November. According to Google’s internal benchmarks, the model achieved a score of 77.1% on ARC-AGI-2, a rigorous test that evaluates a model's ability to solve entirely new logic patterns it hasn't seen during training. This represents more than twice the performance of the 3.0 Pro model and narrows the gap with last week's "Deep Think" specialised mode.

New Capabilities: Code and Creative

Beyond improved logic, the 3.1 Pro update introduces specialised features for developers and designers. A standout innovation is the model's ability to generate website-ready, animated Scalable Vector Graphics (SVGs) directly from a text prompt. Unlike video files, these animations are built in pure code, meaning they remain crisp at any scale and maintain tiny file sizes.

In a live demonstration, the model also successfully architected a complex 3D "murmuration" simulation where users could manipulate the visuals via hand-tracking, complete with a generative audio score that reacted to the movement in real-time. The model retains its industry-leading 1 million token context window, allowing it to digest entire code repositories or thousands of pages of technical documentation in a single interaction.

Optimisation and "Medium" Thinking

To help manage the trade-off between performance and speed, Google introduced a new "Medium" thinking level. This native API parameter allows the model to pause and reason through a problem more deeply than a standard chatbot, but with less latency and cost than the heavy-duty "Deep Think" mode. Developers can now adjust this logic depth on a per-request basis to optimise for high-throughput tasks or deep bug-finding.

Availability and Access

Starting today, February 20, 2026, Gemini 3.1 Pro is rolling out across Google’s ecosystem for various tiers of users:

• Developers: Available in preview via the Gemini API in Google AI Studio, Gemini CLI, and Google Antigravity, the company’s new agent-first development platform.
• Consumers: Accessible to subscribers of Google AI Pro and Ultra plans within the Gemini app and NotebookLM.
• Enterprises: Integrated into Vertex AI and Gemini Enterprise for businesses building specialised autonomous workflows.

The pricing for the Pro model remains consistent with its predecessor, starting at $2.00 (approx. ₹166) per 1 million input tokens for standard prompts, effectively offering users a significant intelligence upgrade at no additional cost.