China AI

AIMathForm-8B is an open-source autoformalization model from OpenBMB that translates natural-language mathematical statements into Lean 4.

AILongCat-2.0, Meituan's model, is now live on the Nous Portal and free to try with Hermes Agent for one week.

AIDeepSeek has released DeepSeek Harness v0.1 in Developer Preview, opening the codebase under the MIT license for developers building agent harnesses.

Why it matters: The source specifies the MIT license and a plugin-based architecture covering models, tools, and sessions, which helps developers assess extensibility before adopting it.

AIByteDance has released Bernini-Diffusers-v2 on Hugging Face, a video generation and editing pipeline combining a Qwen2.5-VL planner with Wan2.2 diffusion components.

AIDeepSeek is updating its API pricing alongside the V4 lineup release, introducing separate peak and off-peak rates.

AIDeepSeek announced the launch of DeepSeek-V4-Pro, citing major Agent upgrades and flexible reasoning effort settings of low, high, and max for V4-Pro and V4-Flash.

Why it matters: The post lists reasoning effort levels and OpenAI Responses API support, giving developers concrete settings to weigh against their existing Agent and coding workflows.

AIDeepSeek has made DeepSeek-V4-Pro generally available on its app, web, and API, with the API model name set to deepseek-v4-pro.

Why it matters: The update pairs new agent benchmark results with API format and pricing changes, so developers can judge both capability and cost impact before migrating.

AIDeepSeek has released DeepSeek-V4-Pro-0813 as the official version superseding the V4-Pro preview, built on the preview structure with a DSpark speculative decoding module.

Why it matters: The release reports agent benchmark gains over the preview and lists vLLM and SGLang setup, useful for judging deployment cost and fit.

AIMiniMax introduces Music 3.0, a music generation model that composes, arranges, performs, and produces a complete song from a creative concept and optional lyrics.

Why it matters: The post explains how the model's pipeline targets structure, acoustic detail, and vocal realism, which helps readers judge where open-weights music generation stands.

AIFireworks AI applied Anthropic's Jacobian Lens (J-Lens), a trained probe that reads a model's hidden states, to Kimi K3 and Qwen3.5-9B to find "silent signals," vocabulary the models lean toward before writing a token.

AIJazzyear's Insights reports that China's tech investment is regaining momentum in 2026, with venture financing up 52% year on year to 2,865 deals in Q1 and AI funding exceeding 110 billion yuan, up 185.4%.

AIMistral AI is expanding its platform to third-party open models, beginning with Z.ai's GLM-5.2. The move lets enterprises match workloads to the right model while keeping the intelligence they build.

AIa life update: i started a new company called Pragmatik (p7k) Labs (语用科技) in shanghai, focusing on the research of next-generation agents across digital and physical worlds. thanks to Gaorong Ventures and HSG (红杉中国 & 高榕创投) for co-leading this round, and to Tencent (腾讯) and Shanghai Engine Fund (上海未来产业基金) for the support. @pragmatik_labs ·

AISkywork promotes using its platform to turn business ideas into fully designed websites. The post offers no further technical details, figures, pricing, or availability information.

AIMoonshot AI's open-weight model Kimi K3 is now live on Databricks through Unity AI Gateway.

AIZ.ai says its ZCode platform has reached 1 million users, and it has reset usage limits for all GLM Coding Plan users as a thank-you.

AIQwen has released the Qwen3.8-2.4T-A95B model weights on Hugging Face, with 2.4T total and 95B activated parameters in a mixture-of-experts design.

Why it matters: The model card gives parameters, architecture, reasoning controls, and benchmark tables against named rival models, showing what an open release of this scale actually offers.

AIMiniMax Music 3 is a music generation model that creates complete songs up to five minutes long from lyrics and a music description.

AIMoonshot AI's Kimi Vendor Verifier has officially verified vLLM to serve Kimi K3 with full accuracy, according to Inferact. Inferact says vLLM delivers its best results across vision, 1M context, and agentic coding at production quality.

AIShanghai AI Lab open-sourced Mobius, an architecture its authors compare to the RNN-to-Transformer shift in both token and knowledge dimensions.

Why it matters: The post links Mobius to the RNN-to-Transformer shift and gives speed, data, and generalization figures that readers can weigh against standard Transformers.

AIAlibaba's Qwen team has released Qwen3.8-27B on Hugging Face as a 27B dense model with native image and video understanding.

Why it matters: The model card gives per-benchmark comparisons with Qwen3.6-27B and named rivals, plus reasoning_effort and preserve_thinking controls for judging cost and agent behavior.

AIMeituan's LongCat-2.0, a coding-optimized model with 1M context and full open-source availability, is now free to use on OpenCode.

AIthat learns the velocity field of a compact dynamic latent space rather than predicting the next frame.

Why it matters: The paper replaces frame-by-frame prediction with a learned velocity field, a different approach to robot world models that readers building embodied systems can compare with video-prediction methods.

AIThomas Dohmke argues Europe must commit to global, model-agnostic open-source collaboration to catch up in the AI era.

AIMeituan LongCat invites users to try its agentic coding demo videos, each of which includes its prompt, using LongCat-2.0. The post directs readers to the LongCat API platform and a Discord community to share their own creations.

AIMeituan LongCat shared a demo in which LongCat-2.0 converted a written game concept into a playable mini game. The post links a short video of the result but gives no details on the model's design, benchmarks, or availability.

AIMeituan LongCat showed a demo in which a single prompt produced an immersive web landing page from a cinematic concept. The post is a short video teaser and gives no model name, version, or technical details.

AIMeituan LongCat turns one natural-language prompt into a full CG animation sequence, covering code, motion, and rendering. The post is a short demonstration video with no further technical details on the model or its performance.

AIMeituan LongCat shared a demo that turns a text description into a rendered 3D planet, covering geometry, materials, lighting, and presentation. The post is a short video showcase and provides no model name, version, or benchmark figures.

AIMeituan LongCat posted a voxel art video showing a "winter pagoda" constructed block by block without any blocks being placed by hand. The post includes a short-form video link but gives no details on the method, model, or tools used.

AIMeituan LongCat says it has kept building with LongCat-2.0 since its release, showcasing five projects each generated from a single prompt. The examples span voxel art, 3D modeling, CG animation, a landing page, and a mini game.

AIAt WAIC 2026, Shanghai AI Laboratory's Bowen Zhou asked whether external evaluations, red teaming, and third-party verification suffice to grant AI real-world authority, and Nitzberg answered no.

AIAli Ghodsi said Databricks has the fastest and lowest latency for Kimi K3 (max), citing an Artificial Analysis provider comparison page. The post credits the Databricks AI team and links to the provider benchmark results.

AIMeituan LongCat has officially launched its international site, offering overseas users a smoother experience. The API platform and chat interface are available at longcat.ai, with a Discord channel for support.

AIMeituan LongCat is accepting business partnership applications from companies and organizations.

AIKimi Slides handles the full slide-building process, from structure and research powered by Kimi K3 to cohesive design with polished charts and SmartArts.

AIDeepSeek has released DeepSeek-V4-Flash-0731 as the official version superseding the preview, with substantially enhanced agentic capabilities.

Why it matters: The release shows benchmark gains over the preview and a concrete vLLM and SGLang serving path, useful for teams weighing a self-hosted agentic coding model.

AISkywork's first batch of its Skywork Note AI hardware device sold out one week after launch, prompting an accelerated rollout of the wider family, including the recording clip, the Recall pendant, and the TriRing AI ring.

AIDeepSeek says DeepSeek-V4-Flash-0731 keeps the same model architecture and size as the preview version.

AIDeepSeek has released the official DeepSeek-V4-Flash API in public beta, with substantially upgraded agent capabilities.