Google Unveils Gemini 3.1 Pro: New AI Model Dominates Benchmarks Again!

Google continues to push the boundaries in the competitive landscape. On Thursday, February 19, 2026, the tech giant unveiled the Gemini 3.1 Pro, the latest version of its most sophisticated AI model, just three months after the release of Gemini 3 Pro in November. Available for preview through Gemini’s API, AI Studio, Vertex AI, the Gemini app, and NotebookLM, the model maintains a context window of one million tokens and is priced at $2 per million input tokens. This update is provided at no additional cost to existing developers and poses a new challenge to competitors like Anthropic and OpenAI.

Gemini 3.1 Pro: Engineered for Complex Tasks

According to Google, the Gemini 3.1 Pro is tailored for scenarios “where a simple answer won’t suffice”. The model offers deeper reasoning, enhanced capabilities to synthesize large volumes of data, and improvements in code generation, including the ability to create SVG animations directly from a text prompt. Google also highlights more efficient management of reasoning tokens, which enhances the model’s reliability for longer and more agentive tasks. In its blog post, Google notes that the “feedback and rapid progress made since the release of Gemini 3 Pro have facilitated these quick enhancements”.

In terms of benchmarks, the tradition continues: the new model excels. Google claims the top spot in 13 out of 16 assessed tests, with logical reasoning scores reportedly more than doubling those of Gemini 3 Pro. However, its dominance has its limits. Anthropic’s Claude Opus 4.6, which was released two weeks earlier, still leads in several tests, particularly those measuring human expert preference. Meanwhile, OpenAI’s GPT-5.3-Codex remains ahead in specialized coding tasks. In short, Gemini 3.1 Pro sweeps the board on paper, but its rivals are likely to regain the lead with their next updates.

Timely Release in a Rapidly Evolving Market

The announcement is strategically timed. It follows a pattern where each developer sequentially releases new models. Anthropic introduced its Claude 4.6 models in early February, OpenAI responded with GPT-5.3-Codex, and Google has now followed up with this update. This occurs in a setting where the historical dominance of ChatGPT is gradually diminishing.

As we analyzed at the end of January, OpenAI has lost its top position in the LMArena rankings to Google and then Anthropic, while its web traffic slows in comparison to the rapid growth of Gemini. This week, Google has also integrated Lyria 3, its musical generation model, directly into the Gemini app, further widening the gap in terms of multimodal functionalities.

The pace at which AI models are being released is so fast that a model touted as “the most powerful on the market” now has a lifespan of only a few weeks. Gemini 3.1 Pro currently tops most rankings. How long it will hold this position is another question.