AI News

Anthropic has released its latest flagship model, Claude Sonnet 4.5, which it outright declares to be "the world's most powerful coding model". This is not just a regular iteration of the model, but a complete leap forward in the building of AI intelligences (agents), computer manipulation, and complex reasoning capabilities. Generation .....

Ollama has become an indispensable tool for any developer wishing to run open source large language models locally. It greatly simplifies the process of downloading, deploying, and managing models. However, with the explosion in the size of model parameters, even specialized PCs can hardly afford to move...

Microsoft recently announced a major update that officially welcomes Anthropic's family of models to its model library for Microsoft 365 Copilot. This means that Copilot, which has long been driven primarily by OpenAI models, now offers enterprise users a second top-tier model selection...

Google Translate is getting a major update. Leveraging the powerful Gemini artificial intelligence model, Google has introduced new real-time conversational translation features to the app and piloted a language practice tool designed to meet users' individual needs. These updates mark Google...

When Google Chief Scientist Jeff Dean and Amin Vahdat, VP of AI Infrastructure, co-signed a technical report, the entire industry listened. The report's core message was packaged in a highly communicative metaphor: Processing a Gemini model's text instructions once consumes more energy than...

In August 2025, a programmatic document called the "Artificial Intelligence Plus" initiative was issued from Beijing, setting a new tone for China's development direction in the next decade. This is not an ordinary policy announcement, but more like a precise roadmap for action and a clear market guide, signaling China's determination to move AI from...

Recently, OpenAI and biotech startup Retro Biosciences announced the results of a collaboration that demonstrates the great potential of artificial intelligence in the life sciences. Using an AI model, GPT-4b micro, designed specifically for protein engineering, the team successfully reprogrammed induced stem cells...

At the 2025 New Products Conference held on August 20, Enterprise WeChat officially launched version 5.0. The core of this update is to fully integrate AI capabilities into office collaboration scenarios, launching three core functions: intelligent search, intelligent summarization and intelligent robots, aiming to build an integrated intelligent office...

A mysterious AI model called Nano Banana is quietly triggering a storm in the technical circle. It has no formal launch event or overwhelming marketing campaign, but with its amazing performance on LMArena, an AI model evaluation platform, it has quickly become a global developer and creative...

Microsoft CEO Satya Nadella recently announced the introduction of a built-in function for Excel: =COPILOT(). This update encapsulates the power of large-scale language modeling directly into cells, making Excel, a nearly 40-year-old piece of software, once again a distribution vehicle for cutting-edge technology. Unlike many through the side...

Recently, Immersive Translate, a popular browser translation extension, was involved in a serious data breach. A large number of "web page snapshots" containing sensitive user information, including high-value business contracts, personally identifiable information, and even cryptocurrency keys, were found to have been publicly exposed on the Internet...

The Alibaba Qwen team recently released an open source command line tool called Qwen Code, whose most notable feature is that it offers developers up to 2,000 free calls per day. This initiative aims to dramatically lower the barriers to using advanced AI coding tools, allowing a wider range of developers...

Since the release of OpenAI's Sora model, the field of AI video generation has been catapulted to unprecedented heights. This wave of technology has not only demonstrated the amazing potential of transforming text into lifelike video, but has also spawned a global race for innovation. Major tech companies and startup teams have entered the field in an effort to...

Since its launch in March 2023, Microsoft's Bing Image Creator has been an active player in the free AI image generation space. Recently, Microsoft upgraded its service again, announcing the integration of OpenAI's latest GPT-4o model into Bing ...

On August 5, 2025, OpenAI dropped a bombshell by officially releasing two open source language models called gpt-oss-120b and gpt-oss-20b. This move marks OpenAI's return to open source for large-scale language models for the first time since GPT-2, a move widely recognized...

Google recently announced that its latest AI feature, Deep Think, has been officially opened to Google AI Ultra subscribers. The move signifies that Google is exploring a different development path from the current mainstream speed-seeking AI models - "Deep Think". Meanwhile, a stronger...

There has long been a central contradiction in the consumer 3D printing space: printing devices are becoming more and more affordable, but the barriers to 3D modeling remain high. Users spend thousands of dollars on a high-performance 3D printer, only to find that the model they want to print either relies on existing shares from the creator community or requires...

Recently, CodeGeex plugin, an AI programming assistant under Wisdom Spectrum AI, has received a noteworthy update. The update not only includes GLM-4.5, a new generation of models developed by CodeGeex AI, but also accidentally integrates a large model from competitor Anthropic, which is shown in the plugin as Claud...

The Byte Jump Seed team has released an experimental language model called Seed Diffusion Preview, specialized for code generation. The model is based on the discrete state diffusion technique and achieves an inference speed of 2146 token/s, which is faster than an autoregression of the same size (Autoregressi...