Google Gemini 3.1 Pro: Benchmarks, Pricing, and Use Cases
Last week, Google rolled out a major update to Gemini 3 Deep Think, focusing on tackling complex challenges in science, research, and engineering. Shortly afterward, Google officially introduced the core model powering these upgrades—Gemini 3.1 Pro. According to Google, Gemini 3.1 Pro further strengthens the reasoning capabilities and long-horizon task handling of the Gemini 3 family, laying a more reliable foundation for developers building autonomous AI agents.
Yao Shunyu, who participated in the Gemini 3 Deep Think research, also shared the progress on social media and noted that even stronger models are expected to continue emerging.
A Leap in Reasoning: Stronger Results Across Benchmarks
Google claims that Gemini 3.1 Pro sets a higher capability baseline for solving complex problems, with clear improvements shown across multiple evaluations. For example, on ARC-AGI-2—a benchmark designed to measure a model’s ability to solve entirely new logical patterns—Gemini 3.1 Pro achieved a verified score of 77.1%, which Google says represents more than double the reasoning performance of Gemini 3 Pro.
Google also shared internal benchmark results indicating that Gemini 3.1 Pro remains highly competitive across several expert domains:
Scientific knowledge: 94.3% on GPQA Diamond
Coding: Elo 2887 on LiveCodeBench Pro; 80.6% on SWE-Bench Verified
Multimodal understanding: 92.6% on MMMLU
Google emphasized that these advances are not merely incremental gains in parameters or benchmark scores. More importantly, they reflect improvements in how the model handles “thinking tokens” and executes long-duration tasks, suggesting greater stability and controllability in tasks that require multi-step reasoning over extended contexts.
Third-Party Evaluation: Gemini 3.1 Pro Reportedly Takes the Lead
According to an evaluation by the third-party organization Artificial Analysis, Gemini 3.1 Pro has moved to the top of its rankings. The report also noted that Gemini 3.1 Pro scores about four points higher than Claude Opus 4.6, while costing less than half as much to run (depending on usage scenarios and configuration).
From Capability to Deployment: Expanding Across Developer, Enterprise, and Consumer Products
Google is now deploying Gemini 3.1 Pro across its product ecosystem for both developers and consumers, bringing this capability upgrade into real-world usage.
Availability and preview channels include:
For developers: Gemini API via Google AI Studio, Gemini CLI, the agent development platform Google Antigravity, and Android Studio (preview)
For enterprises: Vertex AI and Gemini Enterprise
For consumers: the Gemini app and NotebookLM
Google also said that enterprise partners have begun integrating the Gemini 3.1 Pro preview and have reported noticeable improvements in reliability and efficiency. For instance:
Databricks CTO Hanlin Tang stated that Gemini 3.1 Pro achieved “best-in-class results” on OfficeQA, a benchmark for fact-based reasoning over tabular and unstructured data.
Cartwheel co-founder Andrew Carr highlighted improved understanding of 3D transformations, noting that it addresses long-standing issues in 3D animation pipelines such as rotation-order bugs
Pricing: Tiered Token Costs, Plus Cache and Grounding Fees
Notably, Gemini 3.1 Pro’s pricing structure is relatively complex. Key items disclosed by Google include:
Input pricing:
Prompts ≤ 200K tokens: $2.00 per million tokens
Prompts > 200K tokens: $4.00 per million tokens
Output pricing:
Prompts ≤ 200K tokens: $12.00 per million tokens
Prompts > 200K tokens: $18.00 per million tokens
Context caching: $0.20–$0.40 per million tokens, plus $4.50 per hour per million tokens for storage
Grounding (web search): first 5,000 prompts/month free, then $14 per 1,000 search querie
What Gemini 3.1 Pro Can Do: Visualization, Data Integration, and Creative Execution
Google positions Gemini 3.1 Pro as a tool built to solve problems where “simple answers aren’t enough.” It can explain complex concepts clearly, organize scattered information into structured outputs, and turn creative ideas into real, usable products. Google’s official demos include:
Generating animated SVG graphics directly from text prompts (e.g., the “pelican riding a bicycle” example)
Bridging complex APIs with user-friendly design to build real-time dashboards (e.g., visualizing orbital telemetry data)
Producing advanced interactive designs and 3D simulation code for immersive experiences
Translating literary themes into modern front-end portfolio sites and design implementations
What’s Next: Preview First, Wider Release Later
Google said Gemini 3.1 Pro is currently available as a preview, with further breakthroughs planned in areas such as autonomous workflows, and a broader public rollout expected later. Meanwhile, the Gemini app will gradually offer higher usage limits for Google AI Pro and Ultra subscribers, and NotebookLM is already supporting access for Pro and Ultra users.
Sinokap IT Security Training
In past projects, Sinokap successfully helped numerous corporate clients identify and eliminate phishing emails and malware. These case studies highlight our expertise in addressing information security threats:
1. Phishing Email Prevention
We regularly assist clients in identifying and dealing with several network attacks caused by employees mistakenly opening phishing emails. Through rapid response and blocking of malicious links, we ensure that company data remains secure. Additionally, we provide phishing email recognition training for employees to reduce the occurrence of similar incidents in the future.
2. Malware Removal Quick Guide
Sinokap helps companies quickly clean infected devices, restoring normal business operations. We also conduct regular security drills and training to raise employee awareness of various cyberattacks.
Not only have we helped clients effectively respond to urgent security issues, but we also provide long-term information security solutions. Sinokap’s IT outsourcing services and information security expert team are always by your side, ensuring the safety of your business data and operations.
Table of Contents
Call Us, Write Us, Or Knock On Our Door. We are here to help. Thanks for contacting us!
