China AI

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.

AIHeyGen Video 1.0, a 1080p model post-trained on MiniMax H3, ranks #3 on both Design Arena's Video and Image to Video leaderboards, according to Design Arena.

AIAnt Ling open-sourced five Ming-3.0 speech models spanning TTS, audio understanding, and ASR, from 0.5B to 126B total parameters (6.5B active). vLLM-Omni supports serving them from day 0.

AICline says MiniMax M3.1 Flash matches Opus and GPT on coding at about 25x lower cost, and it is free in Cline now.

AIArtificial Analysis says DeepSeek V4.1 Flash 1017 scores 45 on its Intelligence Index, six points above DeepSeek V4.1 Flash, and ties GLM-5.3 as the #2 open-weights model.

Why it matters: The post places the update against peer models on cost and capability, showing where it sits among open-weights options and what it costs per task.

AIOpenCode says its DeepSeek V4.1 Flash model has been updated to 1017 in OpenCode Go and the Free tier. It adds that no changes are needed on users' end.

AIDaniel Han says Unsloth and llama.cpp support DeepSeek V4.1-Flash-1017 with dynamic GGUF quantizations.

AIAnt Ling says it has open-sourced five Ming-3.0 models across speech synthesis, audio understanding, and speech recognition.

AIUnsloth says DeepSeek V4.1-Flash-1017 now runs locally in a 1-bit quantized GGUF, cutting size from 508GB to 250GB (−52%). The company says the quantized model retains about 88.2% accuracy and needs a device with 128GB+ RAM.

AIMiniMax-M3.1-Flash scores 42 on the Artificial Analysis Intelligence Index, 13 points above MiniMax-M3 and level with GLM-5.3-Flash, but behind DeepSeek V4.1 Flash 1017 at 45.

AIFireworks AI says Muse Spark Open is now available through its Serverless Training API. DeepSeek V4.1 Flash 1017 is added to Serverless training and replaces the prior version.

AISenseTime's Large Device platform, led by CTO Xuan Shanming, now serves 6 trillion tokens per day, and plans 10 trillion per day by Q4 2026.

AIThe first Global Embodied AI Application Innovation Challenge and the 2026 China Robot Competition closed in Hangzhou on October 18, with more than 2,000 teams and nearly 5,000 players competing over three days.

AIDAMO Academy's Xuantie unit releases the Xuantie E908A, which it says is the first RISC-V CPU IP to reach both ASIL-B and ASIL-D, with single-point safety metrics above 90%.

AIZhongxingwei, a Chinese interconnect chip company, unveiled 12 mass-produced chips and three new high-performance interconnect products at a launch event.

AIBaidu's Apollo Go has won the Consumer Service award at the 2026 Self-Driving Industry Awards, held in the UK and organized by CarsOFuture.

AIVercel says MiniMax M3.1 Flash, model ID minimax/minimax-m3.1-flash, is now available free on AI Gateway.

AIApple CEO Tim Cook said on October 15 that Apple Intelligence and Alibaba's Qwen model received approval in China, with beta versions expected later this month.

AIMiniMax has released M3.1 Flash on its API at $0.15 per million input tokens and $0.60 per million output tokens.

AIMiniMax M3.1 Flash reaches #26 of 183 on Design Arena's Overall Frontend leaderboard, a 29-position rise over MiniMax M3 at #55.

AIVercel says MiniMax M3.1 Flash is available free on AI Gateway, with better agentic, coding, and cybersecurity performance. The free offer runs through October 22 when served by MiniMax.

AIAliExpress and Tmall announced the "China Brand Global Growth Plan" on October 19, 2026, at AliExpress's annual summit in Shenzhen, with AliExpress allocating special funds for angel investments in AI hardware brands.

AIErik Kannike releases qwen3-4b-deslop, a Qwen3-4B-Instruct fine-tune trained on a MacBook Air to avoid em dashes, "not X, but Y" constructions and lists of three without a system prompt.

AISenseTime releases a study of GPT-6-Astra's embodied intelligence across 12 simulation benchmarks and 8 real-robot platforms, alongside the open-source SenseNova-RoboRSI harness and a 27B model.

AIMiniMax says M3.1 Flash, previously known as the anonymous Space Bunny model, costs one-third as much as M3 and runs three times faster.

AIMiniMax released M3.1 Flash, which it says reaches leading-model levels on agentic, coding, computer use, and cybersecurity benchmarks.

Why it matters: The post lists benchmark scores against named competitors and gives per-million-token prices, so readers can compare cost and capability directly.

AIInferact says it is an API launch partner for MiniMax M3.1 Flash and will provide the production inference serving stack using its vLLM engineering and Inferact Platform.

AIMiniMax releases MiniMax-M3.1-Flash, a multimodal model that it says improves on M3 across multiple benchmarks, especially complex agentic tasks, coding, and cybersecurity. OpenRouter lists the model as free to use.

AIMiniMax announced M3.1 Flash, which it says was an earlier version called Space Bunny on OpenRouter.

AIMiniMax-M3.1-Flash entered Arena's Code WebDev leaderboard at #12 among open models with 1588 points, up 106 points from MiniMax-M3's 1482.

AIOpenCode says Space Bunny was MiniMax M3.1 Flash, which ranked first on OpenCode for 12 consecutive days and served 106T tokens. The model is now available on OpenCode and free for the next three days.

AIOpenRouter says its stealth model Space Bunny was its top model for 10 straight days, processing 56.7 trillion tokens. The model has returned as MiniMax's M3.1 Flash and is free on OpenRouter for the next 72 hours.

AIMiniMax says M3.1 Flash reaches frontier-level results on agent, coding and visual reasoning tasks at one-third the cost and three times the speed of M3.

AIMiniMax is preparing to release MiniMax-M3.1-Flash, according to its API reference docs, which describe it as a frontier multimodal coding model with a 1M context window and tunable thinking depth.

AILightmake, a 3D printing company founded in April 2025 by Wang Zhiyu, says it is focusing on small businesses and print farms rather than consumers.

AIDeepSeek has merged its Harness, Web, App and Platform teams into a new Product Department led by former Harness head Cui Tianyi. The company is recruiting product designers to improve the consumer experience.

AICITIC Securities says DeepSeek Harness could become the key link that turns DeepSeek from model supply into a platform ecosystem.

AIGuizang (@op7418) jokes that DHH understands the situation well, since DHH uses Tailscale to reach Claude Code from China. DHH replied that he routes all his Claude Code use over his Tailscale network.