Our first release is Gemini 3 Pro, which is rolling out globally starting today. It significantly outperforms 2.5 Pro across the board: 🥇 Tops LMArena and WebDev lmarena.ai leaderboards 🧠 PhD-level reasoning on Humanity’s Last Exam 📋 Leads long-horizon planning on Vending-Bench 2

We’re hiring to build AI agents that can see, think, understand, and generate video — from short clips to long coherent movies xAI. If you live for data, agents, and RL, and have a passion for short/long video generation, come ship the future with us! job-boards.greenhouse.io/xai/jobs/49751…

I’ll be at #NeurIPS2025 from 12/3 to 12/5. Excited to catch up with old friends and meet new ones! We will present our work on RL for LLM latent reasoning: arxiv.org/abs/2505.18454 📍 Poster #312 🗓️ Wed, Dec 3 ⏰ 11 a.m. – 2 p.m. PST Location: Exhibit Hall C/D/E Come say hi!

Excited to present Hybrid Latent Reasoning via RL at #NeurIPS2025! 🧠 ✨ We use HRPO to optimize latent "thoughts" in LLMs, achieving consistent gains on reasoning tasks compared to GRPO and larger LLMs 🗓️ Find us Wednesday at Poster Session 1! 👇arxiv.org/abs/2505.18454

Apple just released the CLaRa dataset on Hugging Face, providing 6M examples for efficient Retrieval-Augmented Generation.

New benchmark from Google Research. Models get better at benchmarks, but do they actually get more factual? Previous evaluations focused on narrow slices: grounding to documents, answering from memory, or using search. A model excelling at one often fails at another. This new

🙏 Thanks Sumit for sharing our work! 🚀 If you are interested in search agents that learn and self-evolve without any training data, check out Dr. Zero. We show how self-evolution can bridge the gap between data-free and supervised search agents. tinyurl.com/dr-zero

Dr. Zero Self-Evolving Search Agents without Training Data huggingface.co/papers/2601.07…

[LG] Dr. Zero: Self-Evolving Search Agents without Training Data Z Yue, K Upasani, X Yang, S Ge... [Meta Superintelligence Labs] (2026) arxiv.org/abs/2601.07055

Super interesting paper from Meta Superintelligence Labs. This work suggests complex reasoning and search capabilities can emerge solely through self-evolution, challenging the assumption that human supervision is necessary for advanced agent abilities. Let's break down the

Super interesting paper from Meta Superintelligence Labs on evolving agents without training data.

Today, we’re introducing Personal Intelligence. With your permission, Gemini can now securely connect information from Google apps like @Gmail, @GooglePhotos, Search and YouTube history with a single tap to make Gemini uniquely helpful & personalized to *you* ✨ This feature

🚀 Big thanks to AK for featuring our work! 🧠 We introduce Dr. Zero: self-evolving search agents that autonomously improves without any training data. By leveraging a self-evolution feedback loop, we enable LLMs to synthesize and learn from increasingly difficult tasks.

"Self-Evolving Search Agents without Training Data" As data gets even more scarce, data-free self-evolution is going to be a hot topic in 2026. And this paper by Meta Superintelligence Labs shows you can get SoTA multi-hop “search + reason” agent with zero human training data.

The Top AI Papers of the Week (January 12-18) - UniversalRAG - Agent-as-a-Judge - Self-Evolving Search Agents - Active Context Compression - Efficient Lifelong Memory for LLM Agents - Extending Context by Dropping Positional Embeddings - Unified Long-Term and Short-Term Memory

My wife told me this morning that my Gemini is smarter than hers. I suddenly realized it’s Gemini’s memory. It’s the first step toward personal intelligence: one brain (model weights), personalized chat. Fun fact: we will never die. We will live in AI’s “memory” forever. I talk

Thanks for having me! 🙏 Looking forward to discussing Dr. Zero and sharing our findings on data-free agentic self-evolution:)

Sam Altman can we start getting free boba if we order through chatgpt

You're telling me that this model, which literally beats Gemini-3 Pro on Artificial Analysis, has just dropped on Hugging Face huggingface.co/zai-org/GLM-5

🚀 Qwen3.5-397B-A17B is here: The first open-weight model in the Qwen3.5 series. 🖼️Native multimodal. Trained for real-world agents. ✨Powered by hybrid linear attention + sparse MoE and large-scale RL environment scaling. ⚡8.6x–19.0x decoding throughput vs Qwen3-Max 🌍201