Just now, Google officiallyreleaselatest Gemini 3.1 Pro Preview version, this is Gemini 3 A major upgrade to the series. This model achieved significant breakthroughs in core reasoning capabilities and achieved results in the ARC-AGI-2 benchmark test. 77.1% The performance is more than twice that of the previous generation Gemini 3 Pro. This launch marks Google’s AI Another important development in the field of large language models (LLM).
Google Gemini 3.1 Pro core capabilities improved
Gemini 3.1 Pro builds on the Gemini 3 series and is designed for complex tasks that require deep reasoning. Google officials stated that this version represents a significant leap forward in core reasoning capabilities, providing a more powerful baseline capability for solving complex problems. According to the data, Gemini 3.1 Pro performed well in a number of key benchmark tests: ARC-AGI-2 reached 77.1%, SWE-Bench Verified reached 80.6%, and HLE (Humanity’s Last Exam) achieved 44% and 51% respectively. These data show that the model has reached industry-leading levels in terms of program code generation, logical reasoning, and professional knowledge understanding.
Practical application demonstration
Google demonstrated a number of practical application cases of Gemin 3.1 Pro in the official release, covering areas such as code animation, complex system integration, interactive design, and creative coding. In terms of code animation, 3.1 Pro can generate website-usable animated SVG directly from text prompts. Because these animations are built with pure code rather than pixel images, they stay sharp no matter how you scale them, and the file sizes are much smaller than traditional video formats.
In complex system integration applications, this model demonstrates the ability to integrate public telemetry data streams and successfully build a real-time International Space Station tracking dashboard. This demonstration highlights the power of 3.1 Pro in bridging complex APIs and user-friendly design.
In terms of interactive design, 3.1 Pro has written a complex 3D starling flight simulation system, which not only generates visual code, but also creates a complete immersive experience. Users can control the flock of birds by tracking their hands and listen to a generative soundtrack that changes as the birds move. This provides researchers and designers with a powerful tool for prototyping sensory-rich interfaces.
In the field of creative coding, when asked to build a modern portfolio website for Roaring Heights, 3.1 Pro went beyond simply summarizing the text. Instead, it deeply understood the atmosphere and tone of the novel and designed a contemporary interface that was stylish and captured the essence of the protagonist.
Launch channels and acquisition methods
Gemini 3.1 Pro is available in multiple channels starting today:
- Developer:Gemini API, Google AI Studio, Gemini CLI, Antigravity agent development platform, Android Studio
- Enterprise users:Vertex AI、Gemini Enterprise
- consumer:Gemini App, NotebookLM (exclusive for Pro and Ultra subscribers)
Google emphasized that the purpose of releasing 3.1 Pro in preview form is to verify the update content and continue to improve areas such as agent workflow before the official full launch. 3.1 Pro in the Gemini App is available today for Google AI Pro and Ultra plan users with higher usage credits.
This week can be described as a “super release week” in the field of large-scale language models. It happens to be the week of Taiwan’s Lunar New Year holiday, and major AI giants have successively released Opus 4.6、Sonnet 4.6、Codex 5.3、Grok 4.2 and Gemini Pro 3.1 and other important models were released at the same time, showing that competition in the AI industry has entered a fierce stage.
Gemini 3.1 Pro API Price Comparison with Competitors
Main model price comparison table (per million tokens)
| Model | Enter price | output price | Total cost (1M in + 1M out) |
|---|---|---|---|
| Grok 4.1 Fast | $0.20 | $0.50 | $0.70 |
| Gemini 2.5 Pro | $1.25 | $10.00 | $11.25 |
| Gemini 3 Pro | $2.00 | $12.00 | $14.00 |
| Gemini 3.1 Pro | ~$2.00* | ~$12.00* | ~$14.00 |
| GPT-4o | $5.00 | $15.00 | $20.00 |
| Claude Sonnet 4 | $3.00 | $15.00 | $18.00 |
| Claude Opus 4 | $5.00 | $25.00 | $30.00 |
* The price of Gemini 3.1 Pro is estimated to be similar to that of 3 Pro (official Preview version price)
future outlook
Google says these rapid improvements are driven by user feedback and the pace of technology advancements since the Gemini 3 Pro was released in November 2024. The preview release of 3.1 Pro is intended to validate updates before full rollout and continue to advance more ambitious agent workflows. Just last week, Google released Gemini 3 Deep Think A major update to address modern challenges in science, research and engineering. 3.1 Pro is the upgraded version of core intelligence that achieves these breakthroughs, and Google is bringing this progress into daily application scenarios.
With the launch of Gemini 3.1 Pro, Google has shown strong catching-up momentum in the large-scale language model competition. Whether it is in terms of reasoning capabilities, code generation efficiency, or multi-modal application integration, this new model provides developers and enterprise users with a more competitive choice. The market will continue to pay attention to its actual performance after the official version is released, as well as its long-term competitive landscape with competing products such as Anthropic Claude and the OpenAI GPT series.
Source: KOCPC Chinese