China AI

AIWorkBuddy adds MiniMax's M3.1-Flash model, which uses 0.08x consumption and supports a 300K context window with max thinking effort. The post invites users to try the model and share what they build in the replies.

AIMiniMax released M3.1 Flash, which it says costs one-third as much as M3 and runs about 3x faster.

AILiquid AI (@liquidai) tells US users to try its d1 models and links to the d1-ultra listing on OpenRouter. The post gives no benchmarks, prices, or specifications.

AIVideoJev-4B, a Qwen3.5-4B-based model from MCG-NJU, scores candidate answers and time spans directly for video question answering and temporal grounding.

AIAnt Group's inclusionAI team has open-sourced Ming-3.0-tts under the MIT license, with a 0.5B dense model and a 7.9B-total, 1.3B-active MoE tiny model.

AIAgiBot says its Lingxi X2 humanoid has completed 46,686 interactions at the outpatient guidance desk of Peking University People's Hospital in Beijing as of September 30.

AITencent SSV is giving teachers across China free access to WorkBuddy's education edition, with 2,000 credits per month for 12 months and no regional or school-level restrictions.

AITencent's AI assistant Marvis is now listed in the Kylin software store for Kylin Desktop Operating System V10 SP1 on x86 and ARM, supporting multiple domestic CPUs.

AIAnt Group releases Ming-3.0-tts, an open-weights text-to-speech model under the MIT License, with a 0.5B dense version and a 7.9B mixture-of-experts variant named Ming-3.0-tts-tiny.

AIOllama says MiniMax M3.1 Flash is free on Ollama Cloud for the next three days, in collaboration with MiniMax.

AIThe Department of Homeland Security ordered all federal civilian agencies to remove DeepSeek products and services, including third-party services built on the app, and bars future purchases.

AIDeepSeek-V4.1-Flash-1017 (Max) scored 1670 on Arena's Code Arena WebDev leaderboard, making it the highest-ranked open-source model there, ahead of Kimi K3 (Max) and 12 points behind GPT-6 Sol (Max).

AIIndigo says the main barrier to agent interoperability is political rather than technical, pointing to walled-garden apps in China and US firms restricting agent access.

AISGLang announced a Low-Latency Mode built with NVIDIA and AMD that raises P90 per-user token speed on DeepSeek-V4.1-Flash.

AIAlibaba's Damo Academy says more than 6 billion chips have shipped based on its XuanTie RISC-V architecture, Chief Scientist Meng Jianyi announced at the RISC-V Summit China in Shenzhen.

AIThe source headline says DeepSeek ranks first and MiniMax second among domestic Flash models.

AITsinghua University's electronic engineering department used five AgiBot G1 robots in an end-of-term exam, with undergraduates controlling them through VR handsets to grasp and fold a jacket.

AIDesign Arena projects DeepSeek-V4.1-Flash-1017 to rank #3 among open-weight models on its Website leaderboard, a 16-position jump past DeepSeek-V4-Flash-0731.

AITinker says DeepSeek has released DeepSeek-V4.1-Flash-1017, which DeepSeek describes as stronger on benchmarks, faster at completing tasks, and shorter in output than the previous version.

AIMiniMax says its M3.1 Flash model is free for all Command Code subscribers for a limited time. Command Code describes the model as offering a 1M token context window and frontier-level coding performance, available on all plans.

AIMiniMax has released M3.1 Flash, a faster and cheaper update to its M3 model, and says agentic tasks cost 70% less to run than on its previous model.

AIDatabricks adds DeepSeek V4.1 Flash 1017 as a new endpoint, with the same architecture and updated weights. The model shows substantial improvements across reasoning, coding, agentic tasks, and visual benchmarks, according to Databricks.

AIMiles Brundage, Georgia Adamson, and Konstantin Pilz report that the US AI lead over China is currently around six months.

AIMiniMax says its M3.1 Flash model, codenamed Space Bunny, is now live on GMI Cloud on day zero.

AIvLLM's Elastic EP scaling now reduces serving downtime to 0.82 seconds when scaling DeepSeek V4 Pro from 16 to 20 GPUs.

AIMiniMax says its final M3.1 Flash model is live and free on OpenRouter for 72 hours. The model returns after Space Bunny, its stealth predecessor, was OpenRouter's top model for 10 straight days and processed 56.7T tokens.

AIMiniMax says its M3.1 Flash model, built for long agentic coding runs where cost adds up fast, is free in Cline from day zero.

AIStepFun hosted a hands-on workshop at LA Hacks 2026 to help participants start building with Step 5 Preview. It also launched the StepFun Challenge, and three teams won.

AIOllama announces day-zero support for MiniMax M3.1 Flash, hosted in the US with zero data retention.

AIMiniMax says its anonymous "Space Bunny" model, an early build of M3.1 Flash, held the No. 1 spot on OpenCode for 12 consecutive days and served 106T tokens.

AIThe Department of Homeland Security has ordered US civilian agencies to remove DeepSeek and DeepSeek-affiliate products, plus third-party products built on DeepSeek implementations, under the Federal Acquisition Supply Chain Security Act of 2018.

Why it matters: The source ties the order to a federal notice and its legal basis, showing how supply-chain security rules can reach AI models even when they are hosted in the US.

AITenstorrent, an inference-focused chip startup, has raised $1.5 billion at a $6.2 billion valuation, more than double its late 2024 valuation.

Why it matters: The round's valuation history and the rejected acquisition interest show how investors now value AI inference chip startups competing with Nvidia.

AIAlexander Doria says Berlin found him in a very Chinese moment of his life. The post gives no further detail about the moment or its context.

AIThe Department of Homeland Security requires all civilian federal agencies to remove DeepSeek products and "any third-party services developed using DeepSeek implementations." Agencies must also stop buying them.

AIModelScope now hosts the MiniMax H3 Character Swap LoRA, an experimental adapter that recasts a selected person in video from an image reference.

AIMiniMax has released M3.1 Flash, a reasoning model that scores 42 on the Artificial Analysis Intelligence Index, 13 points above its predecessor M3, at a cost of about $0.22 per task.

AIOllama says MiniMax M3.1 Flash is available through Ollama Cloud, hosted in the US with zero data retention, with a 1M-token context window and image understanding.

AIDeepSeek-V4.1-Flash-20261017 (Max) is now the top open model on Agent Arena, placing #11 overall with a +5.1% net improvement across more than 7.5K real-world sessions.

AIKling AI has released Kling 4.0, which can generate up to 30 seconds of video in one pass and accepts up to ten keyframe images as input.

AIAmir Efrati says models for chip design are a real thing and points to a DeepSeek and Huawei collaboration. The post gives no further details on the models, their performance, or the scope of the work.