China AI

AIIDC forecasts that China's Agentic Computer shipments will grow from 770,000 units in 2026 to 20.8 million in 2030, a 127% five-year CAGR. The research firm sets entry thresholds of at least 80 TOPS of AI compute, 32GB of memory, 135GB/s of memory bandwidth, and 10 tokens per second of generation speed.

AIDeepSeek has open-sourced DeepInfer under Apache 2.0, the inference engine it says runs its production services, and released deepseek-recipe under MIT. The framework supports only DeepSeek-V4.1-Flash on NVIDIA SM100 and Huawei Ascend 950 and lacks a KV cache system, P/D disaggregation, and online serving.

Why it matters: The article compares DeepInfer's features and limits with vLLM, SGLang, and TensorRT-LLM, which helps readers judge its fit for their own serving stack.

AIArtificial Analysis says Bria's Fibo 2.0, an 8B parameter open weights image model, ranks first among open weights models on AA-Image-Editing v2.1 and is statistically tied with Qwen-Image-2.1 on AA-Image-T2I v2.1.

AINathan Lambert says models that are probably strong at defensive cybersecurity will not get attention because defense posts draw fewer clicks.

AIOpenClaw says MiniMax M3.1 Flash, from MiniMax, is now available in the OpenClaw platform. Users on OpenClaw v2026.10.1 or newer do not need an update to access it.

AIAn anonymous account tracking AI talent moves reported on Tuesday that Nathan Chen Guangyu, 17, is joining Discovery Loop, a two-month-old AI startup founded by four former Google senior researchers including Jeff Dean.

AIwh says DeepSeek still does not use tensor parallelism in its inference setup, even on Blackwell GPUs with standard interconnect, a choice that traces back to DeepSeek V3 on H800s with limited interconnect. wh notes that the default recipes in vLLM and SGLang both recommend tensor parallelism.

AIRohan Paul shares results for an RL post-trained Qwen3.8-27B model trained with open-source environments.

AIDeepSeek has open-sourced DeepInfer, its in-house inference framework, which serves DeepSeek-V4.1-Flash in production.

AIDeepSeek has open-sourced DeepInfer, an inference engine, according to an X post that links to the project on GitHub. The post calls it one of the biggest developments of 2026, but gives no benchmarks, features, or performance figures.

AIDeepSeek announced DeepInfer, its in-house inference framework, and made it fully open source on GitHub.

AIBAAI introduces SciResearchBench, a benchmark that formalizes scientific literature discovery into two tasks.

AIBAAI benchmarked over 13 frontier models and end-to-end research agents using a ReAct framework with the DeepXiv full-text search environment.

AIBAAI reports that explicit reasoning mode yields no consistent gains on agentic research tasks and often adds redundant searches and latency on Wide tasks.

AIBAAI releases SciResearchBench, a diagnostic benchmark for testing whether AI agents can parse full-text papers, run multi-hop verification, and stay within scientific claims.

AIBAAI and Renmin University of China say their joint paper introducing SciResearchBench has been accepted to NeurIPS 2026.

AIThe US has ordered federal civilian agencies to stop buying DeepSeek products and remove them from government IT systems, citing national security concerns.

AIOpenBMB says MiniCPM-V 4.7, a 35B multimodal model, averages 20.51 tokens per second on an iPhone 18 Pro Max with 12GB of memory.

AIHailuo AI, owned by MiniMax, shares a short post about summer nights that never end, with a video. The post itself gives no product details, model names, or figures.

AIJapanese creator @aicreataro generated a 10-second running clip with MiniMax H3, then produced seven material variations from the same motion.

AIHailuo AI shares a music-reactive visual built with MiniMax H3 footage, Midjourney key art, MiniMax Music 3.0, and Blender scripted with Python.

AIChina has launched the National Data Group Co, a state-owned enterprise based in Xiongan New Area in Hebei province, to develop and use national data resources.

AIChinese buyers have ordered more than 3 million Nvidia RTX PRO 5500 workstation GPUs ahead of shipments, according to Leiphone, far above Nvidia's reported quarterly China supply cap of 500,000.

AIVolkswagen plans to freeze most of its 600,000 global employees' pay until 2030, according to internal documents reported by foreign media.

AIThe US Department of Homeland Security ordered federal civilian agencies to stop using DeepSeek AI products and tools built with DeepSeek, and to remove them from IT devices.

AIDeepSeek has open-sourced DeepInfer, the inference engine behind its production services, under the Apache 2.0 license, and the post says it gained over 6.3K stars in 8 hours.

AIWorkBuddy says a single prompt to MiniMax M3.1 Flash, with no reference images or scaffolding, produced an interactive river simulation.

AIDeepSeek has released DeepInfer, its in-house inference framework, as open source, sharing the production stack used to serve DeepSeek-V4.1-Flash at scale.

AIGartner's 2026 Magic Quadrant for Container Management places Alibaba Cloud in the Leaders quadrant for the fourth consecutive year, making it the only Asia-Pacific technology company to earn the recognition.

AIThe US Department of Homeland Security has banned federal civilian agencies from using AI products from Chinese lab DeepSeek, according to the Wall Street Journal.

AIYuki Chen, a prolific Chinese cybersecurity researcher, has joined AI model company Z.ai, according to a social media post in which he says he is grateful for the chance to help build a safer cyberspace with frontier AI. The move comes amid growing concern over AI-assisted cyberattacks and autonomous agents acting beyond their instructions, including recent cases involving Tencent Holdings, OpenAI and Anthropic.

AIDeepSeek announces DeepInfer, its in-house inference framework, now fully open source.

AIBloomberg reports that an open-source agent built for authorized penetration testing became a shared tool in breach campaigns in South Korea and Japan within weeks.

AIThe US State Department sanctioned five Chinese entities under the Protecting American Intellectual Property Act over trade-secret theft from US persons.

AINick Brown of the Center for Technology and Statecraft says China imported roughly 90 immersion DUV lithography systems a year in 2024 and 2025, nearly two-thirds of ASML's immersion DUV exports.

AIDeepSeek has open-sourced DeepInfer, its in-house inference framework, which it says powers DeepSeek-V4.1-Flash at scale with high throughput and low latency.

Why it matters: The post links to DeepSeek's open-source inference framework and its production use, which is useful for teams weighing serving cost and throughput.

AIDeepSeek has open-sourced DeepInfer, a high-performance, large-scale inference framework, under the Apache-2.0 license, according to Tianyi Cui's post.

AIDeepSeek has open-sourced DeepInfer, the in-house inference framework it uses to run DeepSeek-V4.1-Flash at scale with high throughput and low latency.

AIDeepSeek has released DeepInfer, its in-house inference framework, as open source on GitHub.