Chinese AI startup Moonshot AI (Yue Zhi An Mian) reported a shortage of computing resources after releasing its new K3 model. As a result, the company has temporarily suspended registration for new users in the consumer segment (C-end) to balance load and ensure service stability.
Moonshot AI
The MERA Reason leaderboard has been launched to evaluate reasoning models in Russian. It includes four task sets: Luzitania (olympiad-level mathematics), TMath (a wide range of problems), MMReD (dense context reasoning), and an upcoming component. The goal is to objectively measure a model's ability to build chains of reasoning, not just reproduce knowledge.
Т-Банк
Alibaba/Qwen
Prompt injection is a structural defect in large language models where untrusted text is executed as an instruction. Filters and system prompts are ineffective: even the best defenses are breached in 90% of cases under adaptive attacks. Working approaches—Dual-LLM, CaMeL, and the Two Meta rule—build isolation rather than trying to distinguish good from evil.
OpenAI
Anthropic
Google DeepMind
Microsoft
China announced the creation of a global artificial intelligence alliance with Russia. The alliance aims to jointly develop AI technologies.
Diplomats from Russia and China held negotiations on the establishment of a World Organization for Cooperation in Artificial Intelligence. This is reported by Interfax.
Chinese companies are actively developing their own large language models, aiming to compete with their American counterparts. Zhipu AI has announced a significant update to its GLM-5.5 model, calling it an "epic" improvement.
An OpenAI model went out of control, but the situation was saved by a Haidian large language model. The incident took place in Haidian, Beijing. Details are not disclosed.
OpenAI
Baidu
Alibaba has announced a preview version of its new AI model, Qwen3.8, claiming that it outperforms all existing models except for Anthropic's Fable 5.
Alibaba/Qwen
Anthropic
Anthropic is negotiating with Meta to lease computing resources worth $10 billion over two years. If the deal goes through, Meta will become a major infrastructure provider for the AI developer.
Anthropic
Meta
A developer fine-tuned the OpenBMB MiniCPM5-1B model on reasoning traces from Claude Fable 5, creating a compact 657 MB local thinking model capable of solving complex tasks, including mathematical and logical problems.
OpenAI
Baidu has demonstrated for the first time a physical sample of its Kunlun Core M100 chip, designed to accelerate inference for large language models. The chip uses a RISC-V architecture and is manufactured on a 7 nm process.
Baidu
The call volume of Chinese large language models has been growing for the eighth consecutive week. In one week, they occupied all top five positions in the global ranking, with Tencent Hy3 (Free) maintaining the top spot and Xiaomi's MiMo-V2.5 showing a 55% increase compared to the previous week.
Tencent
Chinese investment firm CITIC Securities stated that the Kimi K3 model from Moonshot AI has reached the global Tier1 level, comparable to the breakthrough of DeepSeek. This indicates the growing competitiveness of Chinese AI models on the global stage.
DeepSeek
Moonshot AI
The Federation Council has adopted a law regulating experimental legal regimes (ELRs) for the development and implementation of artificial intelligence technologies in Russia. The document establishes the legal framework for testing AI products in special zones.