Hangzhou-based tech company, Qunhe Technology, has again garnered industry attention for its open-sourced spatial understanding model, SpatialLM, which was acknowledged by Google in a recent research paper. This model's innovation lies in its ability to enable robots to understand the geometrical relationships of the physical world through ordinary videos, marking a significant breakthrough in robot training. SpatialLM's core function is to convert videos captured by a phone into three-dimensional spatial layout information. Users simply need to record their home layout using their phone, and SpatialLM will...

Mistral has released Mistral Small 3.1, an open-source large language model boasting a 128K context window and superior parameter efficiency compared to GPT-4o Mini.

According to the Science and Technology Innovation Board Daily, Alibaba Cloud has launched a new initiative called "Project T" to accelerate the development of next-generation AI technologies. The project will focus on several cutting-edge areas, including AI engines, large language models (LLMs), and multi-modal technologies. The goal is to meet the growing market demand through breakthroughs in these technologies. The launch of "Project T" signifies Alibaba Cloud's further deepening of its AI strategy. Insiders reveal that this project will not only speed up AI R&D but also attract more top talent.

In the fiercely competitive AI landscape, a million-dollar experiment is quietly revolutionizing large language model (LLM) training. Jieyue Xingchen's research team recently released groundbreaking findings. By utilizing nearly 1 million NVIDIA H800 GPU hours, they trained 3,700 models of varying sizes from scratch, accumulating a staggering 100 trillion tokens. This revealed a universal scaling law dubbed 'Step Law,' offering a novel guide for efficient LLM training.

Large Language Models (LLMs) are constantly evolving in the field of artificial intelligence. Researchers from Carnegie Mellon University (CMU) and HuggingFace recently introduced a new method called Meta Reinforcement Fine-Tuning (MRT). This method aims to optimize the computational efficiency of LLMs during testing, particularly excelling in solving complex reasoning problems. Studies show that existing LLMs struggle with...

The LLMs.txt Generator has received a major update with the release of version 2. This tool quickly converts any website content into text files usable by AI agents or Large Language Models (LLMs), greatly benefiting AI application developers and users. Developed by the @firecrawl_dev team and fully supported by their official llmstxt endpoint, the new version boasts an incredible 10x speed improvement over its predecessor. The LLMs.txt Generator v2...

Tired of hefty monthly OpenAI Operator subscription fees? Nanobrowser offers a powerful solution. It's a completely free and open-source tool, eliminating subscription costs entirely. Simply install the extension, configure your own LLM API key, and enjoy top-tier web automation capabilities. This 'bring your own lunch' approach is not only cost-effective but also provides complete cost transparency, putting you in control of your AI.

Firecrawl's new LLM.txt API allows users to generate an LLM.txt file for any website simply by providing the URL. This simplifies the process of extracting and utilizing website data for large language models.

Alibaba has open-sourced a new inference large language model, QwQ-32B. Benchmarks show performance comparable to DeepSeek-R1, but with significantly reduced VRAM requirements.

Yale University researchers have developed MindLLM, an AI model capable of translating fMRI brain signals directly into text. This groundbreaking technology bridges the gap between brain activity and human language, opening new possibilities for communication and understanding the human brain.

xAI has announced that its latest AI model, Grok-3, has achieved a remarkable feat, topping the Chatbot Arena leaderboard. The model, internally designated "grok-3preview-02-24", demonstrated superior performance across key areas. xAI's Grok-3-Preview-02-24 edged out GPT4.5-Preview by a narrow margin of 1 point. Grok-3 garnered over 3,000 votes and achieved an overall...

According to Bloomberg reporter Mark Gurman, Apple is facing challenges in rebuilding Siri to adapt to the era of generative AI. The company is not expected to release a truly modern, conversational version of Siri until iOS 20 in 2027. However, this doesn't mean there won't be updates to Siri before then. A new version of Siri is expected to launch as early as May, integrating all the AI features Apple has released in the past year.

A revolutionary technology is quietly emerging in the field of artificial intelligence. Inception Labs recently announced the Mercury series of diffusion large language models (dLLMs), a new generation of language models designed for fast and efficient high-quality text generation. Compared to traditional autoregressive large language models, Mercury boasts up to 10x faster generation speeds, achieving over 1000 tokens per second on an NVIDIA H100 GPU. This speed is...

At the 2025 Global Developer Pioneer Conference, SenseTime announced the launch of its open-source low-code platform, LazyLLM, aimed at lowering the barrier to entry for AI application development. Developers can now build complex multi-agent large model applications with approximately 10 lines of code. LazyLLM not only helps users rapidly develop AI applications but also simplifies the development process through intuitive operations, making it as easy as building with blocks. The name "LazyLLM" reflects SenseTime's goal of enabling effortless AI application development.