#China

AIDeepSeek says its more efficient V4.1-Flash architecture lets it serve more users at lower cost and pass the savings on through lower API prices. Peak/off-peak pricing continues, with off-peak rates at 50% of peak rates. The new pricing takes effect at 04:00 UTC on September 10, 2026.

AIDeepSeek says its V4.1-Flash model needs only 1/4 the HBM and 1/8 the SSD storage for its KV cache compared with the previous generation. Because cache-hit charges often make up a large share of agent costs, the company says the compressed cache significantly reduces those costs.

AIAt Arm Everywhere China on September 8, 2026, Arm launched products spanning data center CPUs, mobile compute subsystems, and robotics platforms. The article says CSS for Mobile 2 integrates CPU, GPU, and neural accelerator for agent AI on phones, and Arm's Neoverse CSS N4 and AGI CPU target agent sandboxes in data centers. It also reports that Arm's Total Design ecosystem now covers over 80 partners for physical AI.

AIBaidu has launched AI, Evolving, a new podcast series, with its first episode examining AI's growing role in scientific discovery through Famou's work on pine wilt disease. The post frames this as part of a broader trend in which AI takes on more of the research process itself. It asks whether research agents could become part of the infrastructure of discovery.

AIBAAI and Peking University researchers released official weights for ConsiSpace, a geometry-consistent multimodal framework for spatial reasoning in long-form visual observations. The model is described in the paper "ConsiSpace: Learning Geometric Consistency Matters for Video Spatial Reasoning" (arXiv:2607.17599).

AIZhipu (Z.ai) reported first-half 2026 revenue of US$142 million, up nearly 400% year on year, driven mainly by its open platform and API business. As of end-August 2026, ARR reached US$1.6 billion, and gross margin for the platform and API business rose from negative 0.4% to 24.6%. The company also said its platform had over 7.4 million registered users and that GLM-5.3's gains came partly from a several-dozen-fold expansion in data environments.

AIBaidu and IFAW have launched AI Guardian, a platform powered by ERNIE models that builds on a collaboration since 2020 that has helped remove more than 18,000 listings linked to the illegal wildlife trade. The platform aims to make this wildlife protection technology more accessible, and users can sign in through ai4wcp.com to help protect wildlife.

AIBAAI and Hao Tang's team at Peking University's School of Computer Science released GaussianMind, an audio-language model built on RoboBrain and OpenMOSS Audio. The model is available on Hugging Face and loads through custom Transformers code with trust_remote_code=True.

AIBaidu's latest AI Pulse highlights how its product portfolio, including DuMate, GenFlow, MeDo, and ERNIE Assistant, is built to take AI work from request to finished result. The post also notes that Baidu has become a dual-primary listed company and that its Apollo Go robotaxi service has expanded in Dubai and Hong Kong.

AIHelical Fusion CTO Wei Xishuo told the NFEC2026 AI-for-fusion forum that the company uses an autoencoder to compress hundreds of stellarator shape parameters into three. The company says this lets it predict zonal flow residuals and turbulent transport from more than 15,000 global simulations and generate new configurations with up to 100x better confinement in simulation.

AISGLang offers day-0 support for GLM-5.3 on NVIDIA Blackwell and Hopper and AMD MI300X, MI325X, and MI355X GPUs, using the same runtime and flags. SGLang is also the rollout engine in Slime, the framework Zhipu used to post-train GLM-5.3, so the runtime that generated the RL trajectories now serves the model.

AIUnsloth announces that Qwen3.8-Flash-Next can be run locally through its GGUF quantizations. The source says the 1-bit version needs 75GB of RAM or unified memory, and that the 125B MoE model is reported to outperform Claude-Opus-4.6 (Max).

Why it matters: The source gives concrete local hardware requirements, quantization sizes, and a guide, showing how a 125B MoE model can run on a 75GB RAM setup.

AIDeepSeek V4 Pro 0813 recorded zero refusals across 840 adversarial security tasks in CyberGym testing, solving them at about half the cost per success of the top-scoring model tested, Kimi K3. In the 697-task common cohort, V4 Pro reached a 53.7% reward rate at $2.50 per solved task, versus 47.6% and $9.64 for GPT-5.5 and 5.9% and $33.28 for Claude Opus 4.8.

AIZ.ai released GLM-5.3-Flash on Hugging Face, the first natively multimodal model in the GLM-5 series, with 320B total parameters and 18B active parameters. The source says it outperforms GLM-5.2 across benchmarks at one-tenth the price and approaches Claude Opus 4.8 on coding and agentic benchmarks. It adopts a hybrid sparse and linear attention architecture to reduce long-context serving costs.

Why it matters: The release shows a hybrid sparse and linear attention design aimed at cutting long-context serving costs, which is useful for comparing efficiency trade-offs.

AIZ.ai released GLM-5.3 on Hugging Face, built on the same base model as GLM-5.2, with all gains coming from post-training. The source reports a 50% improvement over GLM-5.2 on Z.ai Code Bench and open-source SOTA on Terminal Bench 3.0 and Agents' Last Exam, with a benchmark table comparing it against Kimi K3, DeepSeek-V4 Pro-0813, Qwen3.8-Max, and others.

Why it matters: The source gives benchmark tables against GLM-5.2 and rival models, showing where the post-training gains concentrate in coding and cyber tasks.

AIThomas Dohmke argues Europe must commit to global, model-agnostic open-source collaboration to catch up in the AI era. He says that approach would let a breakthrough in Beirut benefit Berlin, linking to a Vox piece on Chinese open-weight AI models such as Kimi and DeepSeek.

AIThe 2026 Global Developers Pioneer Conference (GDPS) and International Embodied Intelligence Skills Competition has opened global calls for scenario challenges, teams, and experts ahead of its October 23–25 event at Shanghai Zhangjiang Science Hall.

AIMoonshot AI's Kimi K3 is available from day zero through Baseten's Model APIs, offering fast and reliable access. Baseten is named as the launch partner for bringing K3 to more users.

AIMoonshot AI's Kimi, in collaboration with kvcache-ai, has open-sourced AgentENV, a distributed system for running agent environments at scale. Its components power agentic RL training for Kimi K3, supporting fast snapshot, resume, and fork for large-scale parallel agent workflows. The project is available on GitHub at

AIMoonshot AI is releasing the model weights and technical report for Kimi K3, a 2.8T-parameter MoE model with native visual understanding and a 1M-token context window. The post says the new architecture delivers 2.5x the intelligence per unit of compute, and the company is also opening high-performance attention kernels, an MoE communication library, and infrastructure for running agent environments at scale.

Why it matters: The source names the model size, context window, and released weights, which helps readers compare its scale and openness with other frontier releases.

AIKimi announced Kimi K3, a model whose open weights are coming soon. The post provides no further details on architecture, parameters, benchmarks, or release date.

AIThe essay argues that Western companies increasingly rely on Chinese open-weight models like Qwen and Kimi for post-training, while Western labs cannot lawfully distill from American frontier models. It says Qwen's share of new open-model fine-tunes rose from 1% in January 2024 to 69% by February 2026, citing ATOM's Report. The authors propose controlled teacher access and tighter enforcement against foreign distillation as a domestic alternative.

AIBAAI has released AREX-Base, a 122B-total, 10B-activated Mixture-of-Experts deep research agent built on Qwen3.5-122B-A10B with a 262,144-token context. The model uses an inner research loop and an outer self-improvement loop, and the source reports it scoring 82.5 on BrowseComp and 85.4 on GAIA, under Apache 2.0.

Why it matters: The release pairs a 122B-parameter deep research agent with benchmark tables against frontier and open models, letting readers compare its search-agent results directly.

AIBAAI's AREX-Turbo is a dense 4B deep research agent built on Qwen3.5-4B with a 262,144-token context length. It scores 70.7 on BrowseComp, 81.6 on GAIA and 40.6 on HLE with tools, versus 82.5, 85.4 and 52.4 for the 122B AREX-Base. The model is released under Apache License 2.0 and targets lower-cost research-agent deployment.

AIA Chinese company has revealed an autonomous robot toilet that drives itself to users when summoned by voice command or remote control. The post provides no further technical details, pricing, or availability.

AIKimi has temporarily paused new subscriptions after Kimi K3 demand pushed its GPU capacity close to limits over the past 48 hours. Existing subscribers are unaffected, and compute is being prioritized for current members while new subscription spots reopen in batches. The company also plans to split membership into Kimi Membership for Kimi Web, App, and Work, and Kimi Code Membership for coding workflows.

AIMoonshot AI announced Kimi K3, a native multimodal model with 2.8 trillion parameters and a 1 million token context window. The announcement cites up to 6.3x faster decoding in million-token contexts and about 25% higher training efficiency, and says open weights arrive by July 27, 2026. The author, Soumith Chintala, reposted it with a brief note of congratulations.

AIThe post argues that China's commercial-scale EUV machines should be forecast for the late 2030s and immersion DUV for the mid-2030s, using ASML's development timeline as a reference. It also weighs factors that could push these estimates earlier or later, including state funding, espionage, talent flows, and the use of AI in R&D. The authors note that forecasts placing either milestone in the 2020s would need strong justification.

AIBAAI President Wang Zhongyuan announced a shift in AI from token prediction to physical state prediction in the institute's 2026 annual research report. The report unveiled the full-stack WuJie architecture spanning foundation models, autonomous agents, and hardware-software infrastructure, and noted that BAAI has open-sourced over 200 models with global downloads exceeding 1 billion.

AIZ.ai's release notes list GLM-5.2 as supporting 1M lossless context, with improved long-horizon task performance and reduced context drift and goal forgetting. The company says GLM-5.2 achieves open-source SOTA performance on coding and long-horizon task benchmarks. The page also includes the newer GLM-5.3 and GLM-5.3-Flash entries, which are listed above GLM-5.2.

Why it matters: The page lists a dated series of Z.ai model releases, showing how the coding and long-horizon agent line has evolved from GLM-4.5 through GLM-5.2.

AIYann LeCun points to Project Tapestry, hosted by thealliance.ai, as the answer to fears that no credible Western open frontier model will emerge. The post is a short pointer to the project, with no details on models, benchmarks, or availability.

AIBaidu's PaddlePaddle account announced ERNIE 5.1, which it says cuts total parameters to about one-third and activated parameters to about one-half, using roughly 6% of the pretraining cost of models at similar scale. The post reports benchmark results including 99.6 on AIME26 with tools, surpassing DeepSeek-V4-Pro on τ3-bench and SpreadsheetBench-Verified, and ranking #4 globally on Arena Search. ERNIE 5.1 is available through the ERNIE website and Baidu AI Studio Model Playground.

AIAhmad Al-Dahle argues that the most interesting part of DeepSeek-V4 is its bet on efficient ultra-long context rather than its benchmarks. He says this is the precondition for test-time scaling and long-horizon agents, and cites 27% of V3's FLOPs at 1M tokens. The quoted DeepSeek post announces DeepSeek-V4-Pro (1.6T total, 49B active) and DeepSeek-V4-Flash (284B total, 13B active), both open-sourced with 1M context and API access.

Why it matters: The post argues that efficient 1M-token context, not benchmark scores, is the key bet behind DeepSeek-V4's design for test-time scaling and long-horizon agents.

AIFun-CineForge, from FunAudioLLM, is an open-source toolkit with an end-to-end dataset pipeline and an MLLM-based model for zero-shot movie dubbing across diverse cinematic scenes. The team built CineDub-CN, described as the first large-scale Chinese television dubbing dataset, and reports that its model outperforms state-of-the-art methods on audio quality, lip-sync, timbre transition, and instruction following. Inference code and checkpoints were released on March 16, 2026, and the model runs on a consumer-grade GPU.

AIThe Beijing Academy of Artificial Intelligence (BAAI) published its Emu3 multimodal large model research in Nature, which the post describes as the first large-model achievement led by a Chinese research institution in that journal. Emu3 learns from text, image, and video at scale using next-token prediction alone, reaching generation and perception performance comparable to task-specific methods. The authors frame this as a step toward scalable, unified multimodal intelligence systems.

AIZ.ai has launched GLM-Image, an image generation model built on a multimodal architecture that combines autoregressive semantic understanding with diffusion-based decoding. The update improves knowledge-intensive generation and makes text rendering inside images more stable and accurate, suiting commercial design and educational illustrations.

AIMoonshot AI released Kimi K2.5, an open-source native multimodal agentic model built by continual pretraining on about 15 trillion mixed visual and text tokens. The model card reports a 1T-parameter Mixture-of-Experts architecture with 32B activated parameters and a 256K context length, and it lists benchmark results against GPT-5.2, Claude 4.5 Opus, Gemini 3 Pro, DeepSeek V3.2, and Qwen3-VL-235B-A22B-Thinking. Weights and code are released under a Modified MIT License, with API access on the Moonshot platform.

Why it matters: The model card gives a full benchmark table against GPT-5.2, Claude 4.5 Opus, and Gemini 3 Pro, useful for comparing open multimodal agent models.

AIAlibaba's FunAudioLLM released Fun-ASR-MLT-Nano-2512, an 800M-parameter multilingual speech recognition checkpoint on Hugging Face that supports 31 languages, with emphasis on East and Southeast Asian languages. It is trained on hundreds of thousands of hours of speech and is available through the FunASR toolkit. The source's benchmark tables cover the Fun-ASR family rather than this checkpoint, so no checkpoint-specific accuracy figures are reported.

AIA ByteDance research team won the Best Paper Award at the Association for Computational Linguistics (ACL) 2021 for VOLT, a simple and efficient machine translation solution. The team has open-sourced VOLT for developers to use.