AI 新闻 —— July 27, 2026
PulseAugur 当天浮现的 20 条头条故事 —— 综合实验室、论文及开发者社区的信号进行排序。
-
Verizon partners with Google on AI infrastructure, plans edge data centers · 3 sources tracked
Verizon is launching an "AI Connect" initiative to capitalize on the growing demand for AI infrastructure. The telecom company has secured a deal worth over $1 billion with Google to utilize Verizon's dark fiber network for connecting Google's data centers. Additionally, Verizon…
-
Changxin Technology sets A-share single-day trading record over 100B yuan
Changxin Technology has set a new record on the A-share market for single-day stock trading volume, surpassing 100 billion yuan. This surge in trading activity occurred on its listing day, significantly exceeding previous records. The company's stock price also saw a substantial…
-
Anthropic releases Claude Opus 5
Anthropic has released Claude Opus 5, a new iteration of its large language model. The model reportedly performed a task quickly, generating a response in 182 words and making a joke.
-
Mira Murati's Thinking Machines Lab launches Inkling with open weights and honest positioning
Thinking Machines Lab, co-founded by former OpenAI executive Mira Murati, has launched its first model, Inkling. This open-weights system boasts 975 billion parameters and a million-token context window, capable of processing text, images, and audio. Uniquely, the company acknow…
-
Google releases Gemini 3.6 Flash and 3.5 Flash-Lite, deprecating temperature tuning
Google has released two new specialized Gemini models: Gemini 3.6 Flash and Gemini 3.5 Flash-Lite. Gemini 3.6 Flash is an improved, more cost-effective model for general tasks, focusing on reduced output verbosity and better agentic planning. Gemini 3.5 Flash-Lite is designed fo…
-
Anthropic debuts Opus 5 at lower price, no data retention
Anthropic has released Opus 5, a new iteration of its large language model, at a significantly reduced price compared to its predecessor, Fable. This new model reportedly does not require data retention, addressing privacy concerns. The announcement positions Opus 5 as a more ac…
-
Ant Group releases Ling-3.0-Flash, a 124B parameter hybrid AI model
Ant Group's AI division has released Ling-3.0-Flash, a new hybrid reasoning model with 124 billion parameters that activates 5.1 billion parameters per computation. This model aims to match or exceed the performance of models two to three times its size in areas like basic reaso…
-
Alibaba Group tests new AI Agent "Qianwen Office"
Alibaba Group has begun a limited test of its new AI Agent product, "Qianwen Office." Initially available as a local desktop client, the service is slated to expand to web and DingTalk versions. The announcement comes amidst broader tech news, including a significant stock surge…
-
OpenAI's GPT-5.6 launches with large context window and cost-saving features
OpenAI has released its GPT-5.6 model family, featuring three variants including the flagship Sol model. Despite a politically charged rollout involving government vetting, the model offers a substantial 1.05 million token context window. A key feature is prompt caching, which c…
-
Anthropic's Claude Opus 5 ships with dynamic tool changes and improved benchmarks
Anthropic has released Claude Opus 5, maintaining the price of Opus 4.8 while claiming near frontier intelligence and offering significant benchmark improvements. The release includes two beta features: the ability to dynamically change an agent's available tools mid-conversatio…
-
Google overhauls search with AI Mode powered by Gemini · 2 sources tracked
Google has introduced its AI Mode, a significant shift in search functionality announced at Google I/O 2025. This new mode, powered by Gemini, replaces traditional search results with a conversational interface designed to reason through complex queries, ask clarifying questions…
-
Chinese Academy of Sciences launches AI "emotional intelligence" system
Researchers at the Chinese Academy of Sciences' Institute of Computing Technology have introduced the "Zing" social-cognitive model system, aiming to imbue AI with "emotional intelligence" and trustworthiness. This system includes a benchmark called SoMBench to evaluate AI's und…
-
Anthropic releases Claude Opus 5, Sonnet 5, and Haiku 4.5
Anthropic released its latest Claude Opus 5 model on July 24, 2026, with pricing set at $5 per million tokens for input and $25 per million tokens for output. The update also includes Sonnet 5 and Haiku 4.5 models, suggesting a tiered approach to their AI offerings. The article …
-
Anthropic's Claude AI enables chat history search
Anthropic's Claude AI now allows users to search their chat history, enhancing the conversational experience. This new feature aims to make it easier for users to revisit past interactions and find specific information within their conversations with Claude.
-
Anthropic launches cheaper Opus 5; AMD challenges Nvidia with ROCm
Anthropic has released its latest model, Opus 5, which is priced at half the cost of its predecessor, Fable. This new model also offers the advantage of not requiring data retention. Additionally, AMD is making strides to compete with Nvidia's dominance in AI accelerators by enh…
-
AI agent achieves same answer with 20x fewer tokens by skipping information gathering
An AI agent was tested on a task that typically requires extensive "detective work" to gather information. By employing a novel approach, the agent was able to bypass this information-gathering phase. This optimization resulted in a significant reduction in token usage, achievin…
-
Build a local AI Q&A system using RAG and MCP
This guide details how to build a local question-and-answer system capable of querying personal documents like PDFs and spreadsheets. It leverages Retrieval Augmented Generation (RAG) combined with the Model Context Protocol (MCP) to create a modular pipeline. The system will co…
-
Anthropic releases Claude Opus 5, emphasizing alignment and reduced cost
Anthropic has released Claude Opus 5, a new AI model that is reportedly more aligned and costs half as much as its previous top-tier model. The company emphasizes that the key advancement lies in trust and alignment, rather than just speed. This release positions Anthropic to po…
-
Developer builds fast, cheap prompt injection detector bypassing LLMs
A developer built a prompt injection detector using a deterministic approach with TF-IDF and logistic regression, rather than relying on large language models (LLMs). This method offers significant advantages in terms of latency, cost, and determinism compared to LLM-based detec…
-
Developer shares strategy for effective LLM test case generation
An LLM developer found that automatically generated test cases for their agent were largely ineffective due to a lack of rigorous filtering and a tendency for the generated tests to mimic the style of the generating model. To improve test quality, the developer implemented a mul…