Headlines · Hot List · Forum · 中文

EastVoice

PHYSIX FRONTIER EASTVOICE

2026.11.14 · Saturday

Issue 080

FIVE MINUTES A DAY, THE GLOBAL AI SHIFT

¥809

Unitree share price on 10 November, up from ¥404

4,600 units

AITO M8 orders within two hours of launch

89 percent

US enterprise open-model spend going to Chinese labs

58.0 percent

Best spatial reasoning accuracy among 21 models tested

Top Stories 3 · dual commentary | Section Picks 24 · one-line takes

TOP STORIES

DeepSeek opens the whole app as an API, Tencent's Hy Realtime voice model ships inside Yuanbao, and a 4B-parameter concept world model from Westlake takes six benchmarks.

1Apple spent a decade on humanoids and shipped with Unitree's legs

Apple's HALO humanoid research, demoed with a robot crawling under tables, shifting boxes, squeezing sideways through a half-open door and sitting down on a chair, leans on hardware that traces back to Unitree. No product has been announced.

What it means for you|Apple's humanoid research leans on Unitree hardware, but no product has been announced, so readers see a demo rather than something they can buy.

Editor's take|Apple's contribution here is the policy, not the body. Legs are a purchase order; squeezing through a doorway without breaking the frame is not.

Source: Huxiu (2026-11-13)

2The stock that doubled in twelve days is not Unitree

Unitree ran from ¥404 on 15 October to ¥809 on 10 November, a market cap above ¥320 billion. Shangwei New Materials did better, climbing from ¥103.8 on 23 October on humanoid supply contracts.

What it means for you|Unitree's rise from ¥404 to ¥809 is the headline, while Shangwei New Materials climbed faster on humanoid supply contracts, showing money rotating into suppliers.

Editor's take|Unitree is the story, the supply chain is the trade. When the OEM's multiple gets rich enough, money rotates into whoever sells it skin and joints.

Source: Huxiu (2026-11-13)

3Baidu Cloud signs Pudu Robotics to train embodied models on real site data

Under the deal signed at Baidu World 2026 on 12 November, Baidu Cloud supplies training infrastructure and cloud services while Pudu runs its own embodied foundation-model work, builds multi-scenario data sets and upgrades its agent platform.

What it means for you|Baidu Cloud supplies training infrastructure and cloud services while Pudu runs its own embodied foundation-model work, builds multi-scenario data sets and upgrades its agent platform.

Editor's take|Service robots already work in kitchens and hotels, which makes their footage worth more than any lab demo. This is a cloud vendor buying a data moat by partnership rather than by cheque.

Source: Leiphone (2026-11-13)

SECTION PICKS

24 AI stories worth your time, each with a one-line take.

Humanoid RobotsHikvision's robot unit finishes shareholding reform and files for IPO guidance

Hikvision said on 13 November that Hikvision Robotics has completed its joint-stock conversion and filed listing guidance with regulators, without giving a timetable or size.

What it means for you|Hikvision Robotics has completed its joint-stock conversion and filed listing guidance with regulators, without giving a timetable or size.

One-line take|Another component maker heading for the A-share window while humanoid multiples are still open. Timing is the entire strategy.

Source: ITHome (2026-11-13)

Humanoid RobotsChinese robots learn to signal through a wall, success rate goes from 0 to 100 percent

A team reports a multi-robot coordination method that passes intent through walls on far less information, cutting communication volume by 80 percent while lifting task success from zero to 100 percent in the tested setup.

What it means for you|A multi-robot coordination method cuts communication volume by 80 percent while lifting task success from zero to 100 percent in the tested setup.

One-line take|Robot teams fail on coordination long before they fail on grasping. An 80 percent cut in channel traffic is the kind of result that shows up in a factory, not in a highlight reel.

Source: QbitAI (2026-11-13)

Autonomous DrivingAITO's new M8 takes 4,600 orders in two hours, built for L3

Huawei's AITO launched the new M8 on 13 November at ¥379,800 to ¥469,800, on an architecture designed for L3 autonomy with the ADS 5 stack and a new omnidirectional sensing suite. Orders passed 4,600 units within two hours of launch.

What it means for you|The new M8 launched at ¥379,800 to ¥469,800 on an architecture designed for L3 autonomy, and orders passed 4,600 units within two hours.

One-line take|An L3 claim is free until somebody accepts liability. The order book says buyers are pricing that risk at zero.

Source: ITHome (2026-11-13)

Autonomous DrivingVoyah puts an 896-line lidar on a ¥197,900 car

The 2027 Voyah Zhiyin launched on 13 November with Huawei's 896-line dual-path lidar and ADS 5 at a limited-time ¥197,900, as Voyah handed over its 100,000th EV at the same event.

What it means for you|The 2027 Voyah Zhiyin launched with Huawei's 896-line dual-path lidar and ADS 5 at a limited-time ¥197,900, as Voyah handed over its 100,000th EV.

One-line take|Lidar line count has become the spec-sheet arms race. At this price the sensor is no longer the cost problem, the compute behind it is.

Source: ITHome (2026-11-13)

Autonomous DrivingMomenta's founder waited outside Great Wall's headquarters for a test ride

A long profile recounts how Cao Xudong brought a demo car to Baoding in late 2024 to get chairman Wei Jianjun behind the wheel, a visit that turned into a supply relationship, as Momenta fights for a defined slot among automakers building autonomy in-house.

What it means for you|A profile recounts how Cao Xudong brought a demo car to Baoding in late 2024 to get chairman Wei Jianjun behind the wheel, a visit that turned into a supply relationship.

One-line take|Tier-one autonomy suppliers are being squeezed into commodity or acquisition. Founder-level sales is what a closing window looks like from the inside.

Source: TMTPost (2026-11-13)

Autonomous DrivingAmap ships an international AutoSDK so Chinese cars navigate abroad

Amap released an international edition of its in-car navigation SDK aimed at Chinese OEMs exporting vehicles, after China's car exports reached 5.859 million units in 2024, up 19.3 percent year on year.

What it means for you|Amap released an international edition of its in-car navigation SDK aimed at Chinese OEMs exporting vehicles, after China's car exports reached 5.859 million units in 2024.

One-line take|Map data is the part of the export story nobody prices in. Cars ship in weeks, local map licensing takes years.

Source: Leiphone (2026-11-13)

World Models & Physical AIFei-Fei Li's world model opens to everyone, and one sentence builds a Chongqing hillside

World Labs opened its world model to general use, with a hands-on test showing a single text prompt generating a walkable 3D scene of Chongqing's terraced streets, building on the lab's earlier image-to-world work.

What it means for you|World Labs opened its world model to general use, with a single text prompt generating a walkable 3D scene of Chongqing's terraced streets.

One-line take|The demo is the easy half. What decides this is whether the generated geometry survives a robot driving through it.

Source: iFanr (2026-11-13)

World Models & Physical AIA 4B-parameter "concept world model" from Westlake takes six benchmarks

A Westlake University team reports a 4-billion-parameter concept world model that tops six evaluations, arguing that structured concept representations beat raw parameter count on physical reasoning.

What it means for you|A Westlake University team reports a 4-billion-parameter concept world model that tops six evaluations, arguing structured concept representations beat raw parameter count on physical reasoning.

One-line take|Four billion parameters winning on physical reasoning is an argument about how data is structured, and it is a far cheaper argument to win than scale.

Source: Leiphone (2026-11-13)

World Models & Physical AITwenty-one models, one spatial test, best score 58 percent

A Zhejiang University team built a predictive spatial reasoning benchmark and found that of 21 vision-language models tested, the best reached only 58.0 percent accuracy, with specialised spatial models losing to general-purpose ones.

What it means for you|Of 21 vision-language models tested on a predictive spatial reasoning benchmark, the best reached only 58.0 percent accuracy, with specialised spatial models losing to general-purpose ones.

One-line take|Spatial reasoning is the capability robotics needs most, and the ceiling is a coin flip with better odds. Read every embodied pitch against that number.

Source: Leiphone (2026-11-13)

World Models & Physical AIGeoThinker attacks the geometry problem inside spatial reasoning

Sun Yat-sen University's team released GeoThinker, two methods aimed at the twin failures of spatial reasoning models, geometry that goes unused and compute that explodes, with gains reported across standard spatial benchmarks.

What it means for you|Sun Yat-sen University's team released GeoThinker, two methods aimed at geometry that goes unused and compute that explodes, with gains reported across standard spatial benchmarks.

One-line take|The field has converged on one diagnosis, models see geometry and throw it away. Fixing that is cheaper than another trillion tokens.

Source: Leiphone (2026-11-13)

Foundation ModelsDeepSeek opens the whole app as an API, so anyone can rebuild it

On the evening of 13 November, DeepSeek opened what it calls a full-app API. Flipping a parameter in the existing API yields deep-reasoning search, file input and the rest of the chat product, while speech capabilities including OCR, ASR and TTS ship as separate endpoints that can be called alone or combined.

What it means for you|Flipping a parameter in the existing API yields deep-reasoning search, file input and the rest of the chat product, while OCR, ASR and TTS ship as separate endpoints.

One-line take|This is the most aggressive commoditisation move of the month. DeepSeek is selling the product surface, which means the wrapper business just lost its moat overnight.

Source: ITHome (2026-11-13)

Foundation ModelsTencent's Hy Realtime voice model sings, calls tools and ships inside Yuanbao

Tencent Hunyuan released Hy Realtime, a speech model covering understanding, instruction-following and human-like delivery, adding agentic tool calls and singing. It is already live in Yuanbao voice calls and Tencent customer service, with a cloud rollout to follow.

What it means for you|Hy Realtime is already live in Yuanbao voice calls and Tencent customer service, with a cloud rollout to follow.

One-line take|Voice is where Chinese assistants actually get used, and tool calls inside a live call are the hard part. Singing is the demo, interruption handling is the product.

Source: ITHome (2026-11-13)

Foundation ModelsA 2.8-trillion-parameter Kimi K3 runs on discontinued Optane, no GPU at all

A home-lab user reported running Moonshot's 2.8-trillion-parameter Kimi K3 on Intel Optane persistent memory with no graphics card, producing roughly 1.5 tokens per second.

What it means for you|A home-lab user reported running the 2.8-trillion-parameter Kimi K3 on Intel Optane persistent memory with no graphics card, producing roughly 1.5 tokens per second.

One-line take|One and a half tokens a second is useless and the point stands anyway. Memory, not compute, is the binding constraint at the frontier.

Source: ITHome (2026-11-13)

Foundation ModelsMenlo: 89 percent of US enterprise open-model spend goes to Chinese labs

Menlo Ventures' 2026 enterprise AI report found that 89 percent of US enterprise spending on open-weight models goes to Chinese providers, with Kimi at 61 percent, DeepSeek at 15 percent and GLM at 11 percent.

What it means for you|Menlo Ventures found 89 percent of US enterprise spending on open-weight models goes to Chinese providers, with Kimi at 61 percent, DeepSeek at 15 percent and GLM at 11 percent.

One-line take|American enterprises are voting with procurement, not with policy. The export-control debate is running about a decade behind the invoice.

Source: Huxiu (2026-11-13)

Foundation ModelsUnbounded Frontier open-sources Fx-Bio-1031 and beats GPT-6-Astra blind

Chinese lab Unbounded Frontier released Fx-Bio-1031, a life-sciences model that outscored GPT-6-Astra and Opus-5 in blind evaluations, built for on-premise deployment so biological data never leaves the customer's network.

What it means for you|Fx-Bio-1031 is built for on-premise deployment so biological data never leaves the customer's network, after outscoring GPT-6-Astra and Opus-5 in blind evaluations.

One-line take|"The data does not leave the building" is the only pitch that works in biotech. Open weights plus local deployment is how you get past a hospital procurement committee.

Source: QbitAI (2026-11-13)

Foundation ModelsCICC and Ant Group sign up to build financial models together

CICC and Ant Group signed a strategic cooperation agreement on 13 November covering financial LLM technology, AI-assisted wealth management and full-lifecycle client service, pairing CICC's research franchise with Ant's model and platform work.

What it means for you|CICC and Ant Group signed a strategic cooperation agreement covering financial LLM technology, AI-assisted wealth management and full-lifecycle client service.

One-line take|Securities firms cannot fine-tune their way past compliance, so they partner. Watch whether the model ends up client-facing or stays inside research.

Source: Leiphone (2026-11-13)

Foundation ModelsA vertical model for teaching Chinese to foreigners

At the World Chinese Language Conference in Beijing on 13 November, Yufei launched a vertical model for international Chinese education, offering graded feedback and personalised tutoring for learners and multi-turn assistance for teachers and institutions.

What it means for you|Yufei launched a vertical model for international Chinese education, offering graded feedback and personalised tutoring for learners and multi-turn assistance for teachers and institutions.

One-line take|Vertical models live or die on distribution, and this one arrives with a state-backed conference and a curriculum behind it. That beats a better benchmark.

Source: ITHome (2026-11-13)

Foundation ModelsWall Street warns the next round of compute financing gets harder and dearer

A column argues that with Nvidia retreating from its role as the sector's lender of last resort, AI infrastructure financing faces higher costs, tighter terms and more scrutiny, just as three years of commitments come due.

What it means for you|A column argues that with Nvidia retreating as the sector's lender of last resort, AI infrastructure financing faces higher costs, tighter terms and more scrutiny as commitments come due.

One-line take|The end of circular financing is the single biggest variable in 2027 capex. Chips are bought with other people's money, and that money is repricing.

Source: Huxiu (2026-11-13)

AI SoftwareShanghai police: the ¥510,000 hotel bill was AI-fabricated, 15 people arrested

Huangpu police said the viral "¥510,000 dinner bill" from a Shanghai hotel was produced with AI editing tools by an online liquor seller's team to drive attention and sales, with 15 people detained.

What it means for you|Huangpu police said the viral ¥510,000 dinner bill was produced with AI editing tools by an online liquor seller's team, with 15 people detained.

One-line take|The first AI-forgery case most people will hear about involves a liquor salesman, not a deepfake politician. Enforcement is arriving through retail fraud, one case at a time.

Source: ITHome (2026-11-13)

AI SoftwareopenEuler ships an OS built for agentic infrastructure, 22 million installs

At the openEuler Summit on 13 November, the project released an operating system aimed at agentic infrastructure, with cumulative domestic installations expected to pass 22 million.

What it means for you|The openEuler project released an operating system aimed at agentic infrastructure, with cumulative domestic installations expected to pass 22 million.

One-line take|China is standardising the layer beneath the agent, which is where lock-in actually happens. Models churn, the OS stack does not.

Source: ITHome (2026-11-13)

AI SoftwareAlibaba brings Xiami back as an AI music app at ¥30 a month

Alibaba relaunched the shut-down Xiami brand as HappyShrimp, an AI music creation app now on iOS and Android, with membership starting at ¥30 per month.

What it means for you|Alibaba relaunched the shut-down Xiami brand as HappyShrimp, an AI music creation app on iOS and Android, with membership starting at ¥30 per month.

One-line take|Resurrecting a dead brand is cheaper than building one, and the subscription price tells you this is a toy with a bill attached rather than a streaming service.

Source: ITHome (2026-11-13)

AI SoftwareopenJiuwen releases an RSI-native AgentOS where agents rewrite themselves

openJiuwen released an AgentOS built around recursive self-improvement, letting an agent write, install and iterate on its own features instead of waiting for a vendor release.

What it means for you|openJiuwen released an AgentOS built around recursive self-improvement, letting an agent write, install and iterate on its own features instead of waiting for a vendor release.

One-line take|Self-modifying agents are one security review away from being unshippable in an enterprise, and that review has not been written yet.

Source: QbitAI (2026-11-13)

AI SoftwareCoding agents fail at writing tests, and code pass rates are overstated by half

A re-evaluation found coding agents that look strong at generation collapse when asked to write the tests validating their own output, with real pass rates roughly half of what standard benchmarks report.

What it means for you|A re-evaluation found coding agents collapse when asked to write the tests validating their own output, with real pass rates roughly half of what standard benchmarks report.

One-line take|The agents are graded by tests written by agents under the same assumptions. Half the pass rate is what you get when the mirror is removed.

Source: New Zhiyuan (2026-11-13)

AI SoftwareLingrui Zhixin raises hundreds of millions as strategic capital piles in

Shanghai compute-chip startup Lingrui Zhixin closed a Pre-A+ round in the hundreds of millions of yuan, led by Meituan's strategic arm with follow-on from JD-affiliated funds, SenseTime Guoxiang, SAIC's Shangqi Capital and others.

What it means for you|Lingrui Zhixin closed a Pre-A+ round in the hundreds of millions of yuan, led by Meituan's strategic arm with follow-on from JD-affiliated funds, SenseTime Guoxiang and SAIC's Shangqi Capital.

One-line take|That investor list is a supply-chain roster, not a VC syndicate. Every name on it wants a domestic alternative to a chip it currently imports.

Source: Leiphone (2026-11-13)

📖 The full EastVoice issue on the Chinese forum

EastVoice — five minutes a day, the global AI shift
Physix Frontier (Shenzhen) · covering foundation models, AI software, humanoid robots, autonomous driving and physical AI