February 24, 2026 | 10:48 am
TEMPO.CO, Jakarta - Technology company Google has just released the latest version of its Large Language Model (LLM), Gemini 3.1 Pro. The model, launched last Thursday, is available in preview and will soon be released to the general public.
The new model is considered one of the most powerful LLMs to date. Many observers consider Gemini 3.1 Pro a significant leap forward compared to its predecessor, Gemini 3, which was already considered a very capable AI tool when it launched last November.
At the same event, Google shared statistics from various independent benchmarks, including Humanity's Last Exam. The results show Gemini 3.1 Pro's performance is significantly better than the previous version.
Praise also came from Brendan Foody, CEO of AI startup Mercor. "Gemini 3.1 Pro is now at the top of the APEX-Agents leaderboard," Foody said in a social media post, as reported by TechCrunch.
He also said that this achievement demonstrates how quickly agents are improving in real-world knowledge work. Mercor's APEX benchmarking system is designed to measure how well new AI models perform real-world professional tasks.
This launch comes amidst an increasingly heated competition for AI models. Technology companies continue to release increasingly powerful AI models, particularly to support agent-based tasks and multi-step reasoning. In addition to Google, several other major companies have also launched their latest models in recent times.
Gemini 3.1 Pro Advantages
According to its official website, version 3.1 Pro brings the advanced reasoning engine previously introduced in Gemini 3 Deep Think to a wider user base. This model is designed to address problems that cannot be answered simply.
In practice, Gemini 3.1 Pro is capable of summarizing and integrating large datasets into a single picture, creating animated SVGs (Scalable Vector Graphics) directly from text commands, and solving complex and multi-level technical and scientific problems.
Specifically for animated SVGs, the results are generated in pure code, so they remain sharp at various sizes and have a smaller file size than conventional video formats.
In terms of performance, Gemini 3.1 Pro recorded significant jumps in various benchmarks. The model achieved a score of 77.1 percent on ARC-AGI-2, more than double the Gemini 3 Pro. Additionally, it posted 94.3 percent on GPQA Diamond for scientific knowledge, 80.6 percent on SWE-Bench Verified for agent-based coding, and 85.9 percent on BrowseComp for agent-based search. On LiveCodeBench Pro for competitive coding, the model recorded an Elo rating of 2887, surpassing several of its competitors.
Read: Google Gemini Arena Makes Its Debut at Gunadarma University
Click here to get the latest news updates from Tempo on Google News
Google Gemini Arena Makes Its Debut at Gunadarma University
32 hari lalu
To strengthen AI literacy among students and academics, Gunadarma University became the first campus to present Google Gemini Arena at a university.
Apple Partners with Google for Gemini-Powered Siri
40 hari lalu
Apple states that Google's AI technology provides the most robust foundation for Apple.
Gemini Users Can Now Build AI-Powered Mini Apps with Opal
19 Desember 2025
Google has once again made a breakthrough in artificial intelligence (AI) by integrating the experimental "Opal" project into Gemini.
Google Translate Gets Gemini AI for Smarter Translations
14 Desember 2025
Google refers to this update as the 'most advanced translation quality' now available.
Indonesian Users Generate 18 Million Images Daily with Google's Nano Banana
5 Desember 2025
Indonesia is the second-highest country in the Asia Pacific region to produce images from Nano Banana generative AI.
AI and the News: How it Helps, Fails, and Why That Matters
1 Desember 2025
AI is reshaping the news ecosystem in the fields of search, fact-checking and personalised feeds. If used well, it can support journalism.
Family of Suicide Victim Sues OpenAI, Condemns GPT-4o Model
9 November 2025
The lawsuit also accuses OpenAI of accelerating security testing in order to beat Google's Gemini launch.
Google's Parent Company Alphabet Reports US$102.3 Billion in Revenue
1 November 2025
The largest revenue comes from the Google Search and other advertising segments, as well as YouTube ads.
AI Chatbots Fail at Accurate News, Major Study Reveals
22 Oktober 2025
AI chatbots such as ChatGPT and Copilot routinely distort the news and struggle to distinguish facts from opinion, according to major new study.
Google Gemini Poses 'High Risk' to Children, Says Common Sense Media
13 September 2025
Common Sense Media found that while Gemini clearly informs children that it is a computer and not a friend, the system still has loopholes.
:strip_icc():format(jpeg)/kly-media-production/medias/5375405/original/050547600_1759921524-pexels-kelvinocta16-1973270.jpg)
:strip_icc():format(jpeg)/kly-media-production/medias/5327418/original/014991800_1756180726-alexandr-podvalny-TciuHvwoK0k-unsplash.jpg)
:strip_icc():format(jpeg)/kly-media-production/medias/5395215/original/007113100_1761665026-WhatsApp_Image_2025-10-27_at_13.37.11_2916f394.jpg)
:strip_icc():format(jpeg)/kly-media-production/medias/4701786/original/013309600_1703841814-jonathan-borba-Z1Oyw2snqn8-unsplash.jpg)
