AI Daily: DeepMind SIMA 2 Arrives, OpenAI GPT-5.1 API Opens, Gemini Live Gets a Major Update

Daily AI Trend Watch: Major models are competing to upgrade, and more powerful AI assistants and developer tools are reshaping the industry landscape. From DeepMind’s gaming AI to OpenAI’s next-generation model, and Google Gemini’s multiple updates, AI technology is advancing at an unprecedented pace.

Google DeepMind Introduces SIMA 2: An AI Agent That Plays, Reasons, and Learns with You in Virtual 3D Worlds

Google DeepMind has once again brought a surprise, officially announcing the second generation of SIMA — SIMA 2. A year ago, the first generation of SIMA (Scalable, Instructable, Multiworld Agent) was born, a general-purpose AI capable of following basic instructions in various virtual environments. At that time, the birth of SIMA was a key step in teaching AI to translate language into meaningful actions in the 3D world.

Today, the launch of SIMA 2 symbolizes a new milestone for DeepMind on the path to creating general and practical AI agents. By integrating the powerful capabilities of the Gemini model, SIMA 2 is no longer just an executor of commands, but has evolved into a gaming partner that can interact with users. SIMA 2 not only understands and executes human natural language commands, but it can now also think about goals, converse with users, and improve itself over time.

This advancement is a major step towards Artificial General Intelligence (AGI) and will have a profound impact on the future of robotics and AI embodiment.

Learn more about SIMA 2

OpenAI Releases GPT-5.1 API: The Perfect Balance of Speed and Intelligence

Developers, take note! OpenAI has officially launched GPT-5.1 on its API platform, the latest model in the GPT-5 series, designed to balance intelligence and speed in agent and programming tasks. GPT-5.1 can dynamically adjust its thinking time based on the complexity of the task, making it faster and more token-efficient when handling simple daily tasks.

In addition, this model has added a “no-inference” mode, which can achieve faster responses on tasks that do not require deep thinking, while still retaining the top-tier intelligence of GPT-5.1.

Check out the GPT-5.1 developer documentation

Official Guide: A Complete Guide to GPT-5.1 Prompting

To help developers fully unleash the potential of GPT-5.1, OpenAI has also released an official prompting guide. This guide details how to migrate to GPT-5.1, shape the agent’s personality, guide user updates, and how to optimize intelligence and instruction-following capabilities.

This guide is written based on extensive internal testing and collaboration with partners. Even minor adjustments to prompts can bring significant improvements in reliability and user experience.

Read the GPT-5.1 Prompting Guide

Gemini Live Experience Upgrade: More Natural Tone, Multi-language Switching, and Role-playing

Gemini Live has received a major update, making the conversation experience more vivid and natural. The new model has a deeper understanding of tone and nuances, making interactions feel more realistic.

• Try new roles: You can give Gemini a role, accent, or identity. Whether you want to practice for an interview or make daily interactions more interesting, it’s easy to do.
• Multi-language communication: Effortlessly switch between multiple languages and even dialects in a single conversation.
• Adjust speech speed: You can ask Gemini to speed up or slow down at any time.
• Vivid storytelling: Try asking Gemini to tell the history of the Roman Empire from the perspective of Julius Caesar. It can provide a more dramatic narrative and richer character dialogue.

Explore the new features of Gemini Live

OpenAI’s New Research: Understanding Neural Networks Through Sparse Circuits

Neural networks are at the core of today’s most powerful AI systems, but their internal workings remain elusive. OpenAI’s latest research attempts to unveil the mystery of neural networks by training models to think in a simpler, more traceable way. This research aims to give us a deeper understanding of how AI works.

Learn about OpenAI’s neural network research

Google NotebookLM Introduces “Deep Research” Feature and Supports More File Formats

Google NotebookLM has been upgraded again, introducing the “Deep Research” feature, which automates and simplifies complex online research. It’s like your personal researcher, able to synthesize detailed reports or recommend relevant articles, papers, and websites.

At the same time, NotebookLM now supports more file types, including Google Sheets, Word documents (.docx), and can directly add PDFs from Google Drive, making your research workflow smoother.

Experience the new features of NotebookLM

Google Upgrades Gemini CLI User Experience: Terminal Interaction Can Also Be Elegant

Google has made a major upgrade to the user experience of the Gemini CLI, making terminal interaction more stable, intuitive, and visually smoother. This update completely revamps the rendering foundation of the Gemini CLI, eliminating the visual noise common in terminal applications. Now, you no longer have to put up with flickering screens or jumping input prompts; everything has become smoother.

Learn about the Gemini CLI update details