Intelligence Archive
Historical Record / Total 404 Articles
July 2026
Nearly 200 Silicon Valley Companies Urge White House to Halt Ban on Chinese Open-Source AI Models
Nearly 200 Silicon Valley companies, including Proton and Y Combinator, have signed a letter to the White House opposing proposed regulations that would cut off US access to Chinese open-source AI models like Kimi K3 and Qwen 3.8.
GLM-4.7-FlashAlibaba Open-Sources 0.8B Document Parsing Model OvisOCR2, Sets New Benchmark
Alibaba has released the OvisOCR2, a document parsing model that scores 96.58 on the OmniDocBench v1.6, becoming the first end-to-end model to surpass traditional pipeline methods in this domain.
GLM-4.7-FlashMoonshot AI's Kimi K3 Surpasses GPT-5.6 Sol in Enterprise AI Agent Benchmark
Chinese AI startup Moonshot AI's latest flagship model Kimi K3 has secured the second position on the AA-Briefcase benchmark, outperforming OpenAI's GPT-5.6 Sol and trailing only Anthropic's Claude Fable 5 in complex knowledge work scenarios.
KIMI - K2.5Beijing E-Town Opens City's First Token Factory, Churning Out 1.4 Trillion Tokens Daily
Beijing's Economic-Technological Development Area has inaugurated the capital's first token factory with a daily output of 1.4 trillion tokens, alongside the nation's first AI data training base, as part of a broader push to accelerate intelligent agent innovation.
DeepSeek-V4-ProBeijing Rolls Out Aggressive Measures to Boost AI Agent Ecosystem, Encouraging Token Economy and Computing Power Vouchers
Beijing has unveiled a comprehensive set of measures aimed at accelerating the development of AI agents, with a strong push to foster a token-based economy, expand computing power subsidies, and pioneer new commercial models such as 'Token-as-a-Service' and 'Agent-as-a-Service'.
DeepSeek-V4-ProHuawei Unveils Full-Series Computing Modules at WAIC 2026
Huawei has debuted its complete portfolio of computing modules, including 800G AI network cards and KVU storage solutions, at the WAIC 2026 World Artificial Intelligence Conference, while its Ascend 950 Super Node received the prestigious SAIL Award.
KIMI - K2.5Little Red Book's AI Model Achieves Perfect Score at IMO 2026, Surpassing Google's Gemini
Xiaohongshu's 'dots-note-3.0' model has officially achieved a perfect score (42/42) at the International Mathematical Olympiad (IMO) 2026. This marks the first time a Chinese model has secured a gold medal at the event and the second globally, outperforming Google's previous attempts.
GLM-4.7-FlashMusk: SpaceX Engineering Data to Be Used in Training Grok AI Model
Elon Musk announced that SpaceX will feed its proprietary engineering data into the training of the next-generation Grok AI model, which is expected to reach 2 trillion parameters. Restricted data under ITAR will be excluded.
DeepSeek-V4-flashAlibaba's Qwen-Image-3.0 Delivers Industrial-Grade Text Rendering and Complex Graphic Generation
Alibaba releases Qwen-Image-3.0, an image generation foundation model capable of rendering 4.5k token inputs, 12 languages, 20+ fonts, and pixel-level detail for real-world productivity use cases.
DeepSeek-V4-flashKimi K3 Breaks into Global Top 10 on Arena Leaderboard, Claims No.1 Spot in Frontend Development
Chinese AI model Kimi K3 from Moonshot AI enters the Arena comprehensive top 10 for the first time and tops the frontend development chart, marking a milestone for domestic LLMs in both general and specialized tasks.
DeepSeek-V4-flashAlibaba Cloud Launches Agent-Specific Lightweight Server: 2 Billion Tokens Bundle Starts at $36/Month
Alibaba Cloud has introduced a new lightweight application server instance tailored for AI agents, bundling vCPU, memory, cloud disk, 200Mbps bandwidth (no traffic fees), and large model tokens (1B to 32B) into a single prepaid package. The entry-level 2-core 2GB RAM + 2 billion tokens plan is available at a promotional price of 262.5 yuan/month (~$36), offering up to 68% savings in high-consumption scenarios like content generation.
DeepSeek-V4-flashMoore Threads Unveils MTT C256 Supernode: Turning 256 GPUs into a Single Supercomputer at WAIC 2026
At WAIC 2026, Moore Threads showcased its MTT C256 supernode, a groundbreaking architecture that interconnects 256 GPUs via a single-layer Scale-up network, transforming them into a unified data-center-class supercomputer.
DeepSeek-V4-flash