China AI

AIPeking University's Yu Zhaofei team and the Beijing Academy of Artificial Intelligence (BAAI) released Brainμ-Spike, a small convolutional network for spike camera image reconstruction that is paired with the Brainμ model.

AIinclusionAI has published Realtime-Venus on Hugging Face with two 9B checkpoints: Realtime-Venus-Omni for audio-visual interaction and Realtime-Venus-Audio for audio-only conversation.

AITencent's WeVisDoc-4B, fine-tuned from Qwen3-VL-4B-Instruct, converts page images into structured Markdown with LaTeX formulas and HTML tables.

AITencent's WeVisDoc-4B, fine-tuned from Qwen3-VL-4B-Instruct, scores 95.38 Overall on OmniDocBench v1.6 and 75.54 mean Overall across three PureDocBench tracks.

AIKling AI shared a photo trick for combining two images so they meet in the middle. The post gives no details on tools, steps, or results.

AIHiDream.ai launched vivago R1, a content creation agent, globally, with a domestic version upgrade.

AIv0 now supports multiple AI models through Vercel's AI Gateway, spanning frontier, cheap, open, and fast options. Users can pick from models including Claude, GPT, Kimi, GLM, Grok, and DeepSeek to build their apps.

AIIn a Paint recreation test, GPT-5.6 Astra built the image from layered geometric shapes, while Qwen3.8 Max worked pixel by pixel.

AIKling AI took part in TIFF: The Market 2026 on September 10 through an industry panel, networking reception, and on-site showcase.

AIMiniMax Design lets a single brief drive a full production workflow on one canvas, with the Astra agent working in Blender through an official connector to build the scene and camera direction.

AIHailuo AI, owned by MiniMax, promotes its MiniMax H3 model as a way to create an army of characters without casting or catering costs.

AIHailuo AI shared a sample showing MiniMax H3 can draw with a pencil. A quoted post says a user created an image with MiniMax H3 Design from a prompt, calling the result impressive.

AIAt TIFF The Market, Kling AI hosted an expert panel and industry reception with filmmakers and tech leaders. The discussion explored how AI can make film production more affordable and accessible.

AINovita AI has open-sourced Chord, a W4A16 MoE CUDA operator with BF16 activations, INT4 weights and group-32 scales, built for Kimi K2.x serving shapes.

AIThe vLLM team trained a DSpark speculative decoding draft model for Kimi K3, a 2.8T-parameter model, using the Speculators library on GB300 NVL72 hardware.

Why it matters: The post shows how hidden-state extraction and Mooncake transfers let a 2.8T-parameter model's speculator be trained across multiple nodes, a reusable pattern for similar setups.

AIMiniMax H3 with SGLang-Diffusion and VDN-H3 generates 14.4 seconds of 768p video in 9.0 seconds end-to-end after warmup on 8× B200 GPUs.

AIDario Amodei's essay "We Must Pace the Frontier" proposes slowing capability gains, starting with independent evaluators inside AI companies and extending to international coordination including China.

AIRadixArk's SGLang-Diffusion, paired with VDN-H3, generates 14.4 seconds of 768p video in 9.0 seconds on 8× B200 GPUs.

AISenseNova Skills is an open-source agentic office skill suite from SenseTime that combines Deep Research, Data Analytics, PPT Generation, and Infographic Generation.

AIThe Hailuo AI Bestiary contest closes tonight, September 14 at 23:59 PT, with $8,000 cash, 200,000 Credits, and ten Audience Choice awards.

AIBaidu's second AI, Evolving episode compares how companies in China and other markets build AI agents for the workday. It uses DuMate as a case study of what agents need to earn user trust and help individuals and teams do more.

AIIntern Large Models introduces Intern-S2-397B, a multimodal foundation model built for long-horizon scientific research and scientific agents.

Why it matters: The post pairs a new open multimodal model with benchmark tables against named Qwen, DeepSeek, Kimi, GLM, GPT, Gemini, and Claude models, letting readers compare scientific and agentic results directly.

AITencent released Simple-Attention-Sparsification (SAS) gate checkpoints for Qwen3-4B, Qwen3-8B, and Qwen3-14B, which learn to rank and select KV blocks using continuous gates optimized with the language-modeling loss.

AIBaidu's Miaoda no-code platform has upgraded with enhanced AI agents for design, app generation, and testing.

AIMiniMax highlighted open-source community progress on its H3 video generation model, which it built with native stereo audio and multimodal reference control.

AIHailuo AI promotes bringing AI tools into professional 3D production workflows.

AIAt Guotai Junan Securities' 2026 Autumn Conference, SenseTime's Head of Capital Markets Philip Wong laid out three shifts reshaping AI: from single-modal to native multimodal, from token consumption to task delivery, and from single-point models to system-level full-stack capabilities.

AIQwen has open-sourced Qwen-Image-2.1, a unified text-to-image generation and image editing model with 7B parameters in its visual generation component.

Why it matters: The source specifies the 7B visual component, transparent RGBA output, and up to 10 reference images, which helps readers judge its fit for generation and editing workflows.

AIinclusionAI released Step-3.7-Flash-singprobe, an 8.13M-parameter probe that reuses Step-3.7-Flash hidden states to score query intent, response unsafety, and hallucination risk at every generated token.

AIinclusionAI has released Qwen3.8-27B-singprobe, a 10.1M-parameter intrinsic streaming guardrail that reuses Qwen3.8-27B hidden states to score query intent, response unsafety, and hallucination risk at every token.

AIinclusionAI has released Qwen3.5-397B-A17B-singprobe, an intrinsic streaming guardrail built on Qwen/Qwen3.5-397B-A17B that scores query intent, response unsafety, and hallucination risk at every generated token using the base model's hidden states.

AIinclusionAI released SingProbe, a 5.8M-parameter intrinsic guardrail built on openai/gpt-oss-120b that scores query intent, response unsafety, and hallucination risk at every token.

AIFireworks AI reports that DeepSeek-V4.1-Flash scores 74.34% pass@1 on DeepSWE at $0.430 per task, close to GPT-6-Astra's 74.12% at $6.524.

AIThe Cyberspace Administration of China says 124 new generative AI services completed filing in July and August, including 7 mobile on-device services.

AIThe National Cybersecurity Standardization Technical Committee released the Artificial Intelligence Safety Governance Framework 3.0 on September 14 at the opening of National Cybersecurity Publicity Week.

AIJohn Schulman posted that Anthropic's approach to Elon Musk would be nice to see applied with China. The post gives no detail on what that approach involves.

AIShanghai AI Lab's InternLM team released Intern-S2-397B, a multimodal foundation model for scientific intelligence and long-horizon agents.

AIAidan Gomez, Cohere's CEO, sarcastically criticized proposals from the AI "cartel" that would require employee-level access to operations, allow shutdowns on safety grounds, and withhold chips unless China also complies.

AIEpoch AI's newsletter reports that Huawei is far behind Nvidia and is unlikely to catch up this decade due to export controls.

Why it matters: The newsletter bundles several analyses of AI chips, GDP measurement, and benchmarks, so it helps readers scan the research agenda behind each finding.

AIInferact says vLLM now supports DeepSeek v4.1 Flash on day zero across H100, H200, B200, B300, GB200, and GB300 GPUs.