Daily Highlights
- Meta and Fourier jointly build humanoid robot prototype Max
- Apple brings Alibaba Qwen into China's Apple Intelligence
- GPT-6.2 Astra and new Claude flagships leak the same day
- OpenAI cuts funding target to about $20 billion
- NVIDIA unveils next-gen co-packaged optics technology OOI
TODAY'S HEADLINES
[Meta and Fourier Develop Humanoid Robot Prototype 'Max' as the Physical Avatar of Muse] According to The Information on Nov 1, Meta has jointly developed a humanoid robot prototype called "Max" with Chinese robotics company Fourier, and started work on a second-generation prototype, Max 2, in September. Standing about 1.5 meters tall and weighing about 48 kg, the robot is powered by an NVIDIA Thor chip and took three months to build. Meta sees the humanoid as the "physical avatar" of its Muse AI agent, aiming to extend AI from screens into homes; the project remains internal with no sale plans or launch timeline. (Source: Eastmoney)
[Apple Integrates Alibaba Qwen into China's Apple Intelligence, Rolling Out in December] On Oct 31, Apple announced a partnership with Alibaba to integrate the Qwen large model into Apple Intelligence for mainland China, expected to arrive in the iOS 27.3 developer beta on Nov 4 and ship broadly in December. Users will be able to ask Siri questions powered by Qwen, follow up on documents and images, and generate text and images via "Qwen Creation," with Baidu's Ernie also supporting search and vision capabilities. It marks Apple's first use of a domestic model to complete its system-level AI offering in China. (Source: Eastmoney)
NEW PRODUCTS & TECH
[GPT-6.2 Astra and Claude Mythos/Fable 5.5 Surface the Same Day as Flagship Cycles Compress to Weeks] On the evening of Nov 1, a developer debugging Codex found the underlying model had quietly switched from GPT-6 Astra to gpt-6.2-astra, with billing data showing the new base already serving nearly 20% of real traffic. Around the same time, references to Claude Mythos 5.5 and Fable 5.5 were found in Anthropic's source code — the former a hardened security variant for elite cybersecurity and research institutions, the latter supporting a 1M-token context window. Leaks suggest GPT-6.2 skipped a version after GPT-6.1's safety testing fell short; official capabilities remain subject to launch. (Source: Synced/NetEase)
[NVIDIA Unveils Next-Gen Co-Packaged Optics OOI to Break AI Cluster Interconnect Bottlenecks] On Nov 2, NVIDIA introduced its next-generation co-packaged optics (CPO) technology OOI at SEMICON West 2026. Dropping organic substrates, it mounts the optical engine directly on the GPU's silicon interposer, bringing electro-optic conversion closer to the die and enabling more compact, lower-loss interconnects for very large AI clusters; prototype samples are already in hand. Inter-GPU bandwidth is becoming the key constraint on frontier-scale clusters. (Source: Phoenix Tech)
[Zhiyui Releases Embodied Agent Framework Genie Master Letting Robots Learn on the Job] On Nov 2, Zhiyui Robotics released Genie Master, an embodied AI agent framework that lets robots accumulate experience and self-improve while performing real tasks, lowering the barrier from demonstration to deployment. (Source: Sina Tech)
IPO & FUNDINGS
[OpenAI Cuts New Funding Target to ~$20B at a $1.25T Valuation] According to media reports on Oct 30, OpenAI has lowered its new funding target amid an AI stock selloff, more cautious investor commitments and Anthropic's pending IPO. The company now plans to raise about $20 billion at a $1.25 trillion valuation, down from a previous plan of $30 billion at $1.4 trillion. Some investors have trimmed commitments; UAE's MGX is expected to be a major participant, with SoftBank, Sequoia and Coatue also planning to join. (Source: Eastmoney)
[Amazon Secures Up to $80 Billion Loan Facility for AI Infrastructure Purchases] An SEC filing on Oct 30 shows Amazon has set up a loan facility of up to $80 billion for technology infrastructure asset purchases, drawable at its discretion between Q1 2027 and Q3 2028 with five-year terms. As of end-September 2026, Amazon's total spending commitments rose to $883 billion, up 36% in three months. (Source: NBD)
OPEN SOURCE & RESEARCH
[MiniMax M3.1 Flash Weekly Token Volume Surges 690%, Enters Global Top 10 at No. 8] Based on OpenRouter data, total global LLM token usage reached 194 trillion tokens for Oct 26-Nov 1, up 3.19% week over week. Chinese models hold four of the top five spots, with DeepSeek V4.1 Flash 1017 at No. 1; MiniMax M3.1 Flash jumped to No. 8 on a 690% weekly surge, while Xiaomi's MiMo-V2.6-Flash dropped off the chart. (Source: NBD)
COMPANY & PEOPLE
[Volcengine MaaS Revenue Hits $4.8B RMB in H1, Annual Target Raised Above 15B] On Nov 1, it was reported that ByteDance's Volcengine MaaS business generated 4.8 billion RMB in H1 revenue with a gross margin above 55%, becoming ByteDance's largest AI business by revenue; the annual target has been raised above 15 billion RMB. (Source: Cnyes)
POLICIES & DATA
[xAI to Add Default Watermarks to Grok API Text Output] According to industry media on Nov 2, xAI plans to add watermarks by default to text generated by seven Grok API models starting Nov 7, 2026, for AI-content identification and provenance; the exact model coverage and opt-out options remain unconfirmed. (Source: Phoenix Tech)
[AI 'Tokens' Hit E-Commerce Shelves for Singles' Day, Tmall Sales Up 450% on Day One] On Nov 2, AI tokens became a standalone product category on major e-commerce platforms for the first time this Singles' Day, with Tmall first-day token sales surging 450% week over week; DeepSeek, Alibaba Cloud, Zhipu and Kimi all joined, and MiniMax opened its first official flagship store. (Source: Sina Tech)
---
Powered by AOYII.com
