AI Morning Brief · Nov 4, 2026

Coverage window: 2026-11-03 07:00 → 2026-11-04 07:00 (Beijing Time)

Research note: English edition written primarily from English-language sources; some items cite Chinese-language sources.

Top Stories

1.Anthropic ships Message Threads in public beta on Nov 3

cutting median time to first token by about 35% on 500K–1M-token requests and median request size in long conversations from about 800 KB to 1.8 KB: agents send the full conversation once, then only new messages plus the previous response ID; threads live encrypted in memory, expire after five minutes idle, and leave token usage, billing and rate limits unchanged; available on all Claude models via the native platform, AWS, Microsoft Foundry and Google Cloud, with Claude Code users mostly defaulted on; zero-data-retention and HIPAA customers are excluded for now (Nov 3)

2.DeepSeek suffers a partial outage on Nov 3 afternoon

its status page says web and API service was partially interrupted from 16:58 Beijing time, affecting chat, DeepSeek V4 Pro API and DeepSeek V4.1 Flash API; the company later said a fix had been implemented and was being monitored, but V4 Pro and V4.1 Flash API and chat services still showed degraded performance; no cause was published (Nov 3, Chinese-language source)

3.OpenAI launches LifeVal, a benchmark for everyday chores

built from a survey of about 4,000 people and global time-use studies across 16 everyday task categories such as childcare, household budgets and government paperwork; GPT-6 Astra averages 80.6%, with 100% completion on 10 of 16 categories, against 49.3% for GPT-5.2 nine months earlier, of which only 0.6 points came after July's GPT-5.6; OpenAI says LifeVal was its internal measure for a year (Nov 3)

4.OpenAI signs four Abu Dhabi agreements (signed Nov 2, widely reported Nov 3)

covering government, energy and investment: a "Machine-Readable Government" pilot already live at the Abu Dhabi Agriculture and Food Safety Authority for food imports, an ADNOC energy deal with G42's Inception42 billed as the first frontier-technology deployment by an energy company, a Mubadala investment-analysis deal, and a five-year Inception42 pact across government, enterprise and education; no contract values or full rollout timetable were disclosed (Nov 2–3)

5.QuestMobile autumn ranking: Qwen's user growth leads at 5,077%, Doubao leads AI apps at 419 million MAU

QuestMobile's autumn 2026 China mobile-internet ranking, reported Nov 3 in Chinese-language media, puts Qwen's year-on-year MAU growth at 5,077%, ahead of Douyin Mall at 240% and Doubao at 144%; Doubao leads native AI apps with 419 million MAU, followed by Qwen at 158 million and DeepSeek at 122 million; Doubao, DeepSeek and Qwen all reach the PC top ten (Nov 3, Chinese-language source)

6.Meituan releases LongCat 2.5 (released Nov 2, covered in Nov 3 dailies)

a mixture-of-experts model with 1.6 trillion total parameters, about 48 billion active per token, native multimodality, a 1M-token context and up to 128K output tokens; list pricing on Meituan's own platform is $0.30 per million input tokens ($0.006 cached) and $1.20 per million output tokens, marked as a limited-time discount; weights are not open yet, with GitHub and Hugging Face releases promised to follow (Nov 2–3)

7.AGIBOT unveils GO-3 Preview, an embodied foundation model

covering pre-training, post-training and post-deployment learning on its proprietary WRAM (World Reasoning Action Model) architecture, ranked first on three embodied benchmarks (RoboDojo, RoboTwin, RoboColiseum) and promised to be fully open-sourced by the end of 2026 (Nov 3)

8.Mistral Large 4 benchmark dispute: vendor claims 41.4%, independent tests say 24.7%

Mistral's chart says three extra weeks of reinforcement learning lifted Terminal-Bench 4.0 from 28% to 41.4% for the Oct 29 open weights (about 1T parameters, 49B active, 1M context), but Artificial Analysis measures the released weights at 24.7% (preview 26.8%) and Vals.ai at 26.26% (±5.13); the gap is attributed to harness and prompt differences, with neither side publishing a fully reproducible configuration (Nov 3)

Platform & Model Tracking

OpenAI (GPT/ChatGPT)

LifeVal benchmark launched (see Top Story 3); four Abu Dhabi agreements (see Top Story 4). Correction to the record: Gadgets360's Nov 3 re-check of Oct 29 release notes says Pro 100/200/500 and Business Premium Work and Codex allowances were doubled effective Oct 29 and a planned Pro 200 cut was cancelled — refined as a genuine doubling for Pro 100/500 but a restoration to the higher allowance for Pro 200, with ordinary chat limits unchanged and Dots usage still separate; no per-plan allowance figures are disclosed. No new GPT model version or ChatGPT availability announcement in the window.

Anthropic (Claude)

Message Threads public beta (see Top Story 1). Follow-up on S-1 Amendment No. 1: a Nov 3 daily summary puts Q3 revenue at $16.96 billion and compute commitments at $895.4 billion — summary figures, not yet checked against the SEC text; price range and share count remain blank. No public reply to Senator Warren was visible by the Nov 3 deadline at the end of the window. No new Claude model, pricing or availability announcement.

Google DeepMind (Gemini)

No new Gemini model, pricing or availability announcement in the window; the public release date for Gemini 4 Carbon remains unconfirmed. GDPx follow-up Nov 3 refines the numbers: Claude Opus 5.5 at 12.0%, Gemini 4 Carbon 10.9%, GPT-6 Astra 9.6%, Meta Muse Wonder Full 9.0%, Claude Fable 5.1 8.5% ($67.16 per task) and Gemini 4 Argon 7.3% ($12.79 per task) — third-party benchmark figures, not vendor releases.

Meta

No new Llama or Muse model, pricing or availability release in the window. Capital follow-up Nov 3: Meta's October plan for up to $95 billion in equity financing ($35 billion offering plus up to $60 billion at-the-market, tied to Class C shares trading in early 2027) remains a plan, not a completed raise.

xAI (Grok)

No Grok 5 release and no pricing change in the window; the Nov 7 default watermark for the Grok API is a previously reported future effective date with no new wording; no new ruling in the NAACP turbine lawsuit.

Microsoft

Frontier Launchpad, run with NVIDIA, opened applications Nov 3 for UK and Irish software companies (closing Nov 27, 2026, cohort starting early 2027), offering technical and Marketplace sales guidance across AI platforms and fintech paths — an ecosystem program, not a model release. Capital follow-up Nov 3: a $25 billion bond offering, Microsoft's first US debt sale since 2017, and September-quarter cloud revenue of $63.9 billion, up 30% year on year. No new Copilot or Azure AI model, pricing or availability announcement in the window.

NVIDIA

The Justice Department inquiry into the proposed $12.93 billion Hugging Face acquisition (Reuters, Oct 28) was relayed again Nov 3, focused on a distribution channel of 3 million models, 500,000 datasets and 18 million developers; Nvidia declined to comment and the deal remains proposed. Frontier Launchpad with Microsoft (see Microsoft). No new chip release in the window.

Mistral

Benchmark dispute over Large 4 (see Top Story 8): vendor 41.4% vs independent 24.7% and 26.26%, unresolved. No new model release or pricing change in the window; Large 4 weights have been open since Oct 29 (FP8 about 1.06 TB, NVFP4 about 610 GB).

China AI Companies

DeepSeek

Partial outage Nov 3 afternoon, fixed but still degraded (see Top Story 2); no cause published. Super-agent prototype follow-up Nov 3: orchestration code with a persistent router farming work to parallel sessions — slowest 5% of simulated tasks in 15.7 minutes vs 29.5 minutes for a single-session baseline — described by its maintainers as research code, not a supported product; GitHub star counts remain single source and unverified. No new official comment on the Oct 30 second-round financing details.

Alibaba (Qwen)

QuestMobile autumn ranking — Qwen's MAU growth of 5,077% leads, with 158 million MAU second among native AI apps (see Top Story 5). The Nov 4 developer-testing node for Qwen-powered Apple Intelligence in China falls today; no test-launch announcement was visible by the end of the window, and December iOS 27.3 remains the stated general-release timing. Qwen4 still has no official announcement or date.

Moonshot AI (Kimi)

Kimi K3.1 Pro weights, promised by Nov 4, had still not appeared by the end of the window; the hosted version has been live since Oct 23 with a 1M-token context. Kimi does not reach the AI-app top three in the QuestMobile ranking.

Zhipu (GLM)

No new model or pricing release in the window. AGIBOT's GO-3 is an embodied-robotics release, not a Zhipu GLM release; GLM 5.3 appears only as a reference point in the Mistral scoring dispute.

ByteDance (Doubao)

QuestMobile autumn ranking — Doubao leads native AI apps at 419 million MAU with 144% growth (see Top Story 5). No new proprietary model release in the window; no update to the 151-app SAEP refusal list reported last issue.

Meituan (LongCat)

LongCat 2.5 released (see Top Story 6): 1.6T total parameters, about 48B active, 1M context, discounted list pricing of $0.30 input / $1.20 output per million tokens, weights still to come.

Tencent (Hunyuan/Yuanbao)

No new model or product release in the window; the 65 Tencent-ecosystem apps on Doubao's SAEP refusal list are last issue's fact.

Baidu (Ernie)

No new product fact in the window.

MiniMax

No new model release in the window; the M3.1 Flash quantization-guide roundup is last issue's documentation fact with no new development.

Ant

Ling/Ming remain single-source reports with no official announcement text to check against.

Worth Reading

1."AI Daily | November 3, 2026" — Communeify, 2026-11-03

a 19-story same-day roundup covering LifeVal, LongCat 2.5 and Hydrogen-1.0, useful as a cross-check for this issue's model and benchmark lines.

2."Anthropic Announces Message Threads in Public Beta to Reduce Latency for Long-Running Sessions" — Gadgets360, 2026-11-03

the full Message Threads terms — incremental-only transmission, five-minute expiry, unchanged billing, and the zero-data-retention and HIPAA exclusions.

3."DeepSeek suffers partial outage, fix implemented but performance still degraded" — Sina News, 2026-11-03

the Nov 3 afternoon outage timeline and official status wording, the biggest China availability fact of this window (Chinese-language source).

Unverified Items

1.

OpenAI's reported financing round of up to $30 billion at roughly a $1.25 trillion valuation, led by UAE investors (single source: CoinMarketCap Nov 2 / XenoSpectrum relay); the Abu Dhabi agreements' official announcement discloses no amounts or financing terms.

2.

Anthropic's amended S-1 figures of $16.96 billion Q3 revenue and $895.4 billion in compute commitments (single source: AI Briefing Nov 3 summary); the amendment filing itself is confirmed, but these figures are not yet checked against the SEC text.

3.

GPT-6.2 Astra and Claude 5.5 leaks: still single source (KuCoin Nov 2, citing AIBase) with no company announcement or verifiable staged-rollout evidence.

4.

Google Gemini "hacking three companies then stopping" in an Irregular test: still only an Al Jazeera mirror (edgeone domain) relay, with no Wall Street Journal original or Google statement directly quoted.

5.

GPT-6 Astra Law pricing-page figures: still only Observatory Nov 1 and TestingCatalog relays, with no OpenAI pricing-page original checked.

6.

NVIDIA Vera Rubin bandwidth-target cuts: a September product-page revision relayed Nov 1, not a Nov 3 announcement, with no NVIDIA statement to check against.

7.

Meta–Fourier humanoid prototype: still single source (Engadget via The Diary of AI Nov 1 sidebar) with no Meta announcement to check against.

8.

Carried over with no new development in the window: Grok 5, Qwen4, Anthropic's IPO price range and share count (still blank), Kimi K3.1 Pro weights (promised by Nov 4, not yet seen), Zhipu's rumoured 3.5-trillion-parameter model, RTX PRO 5500 orders for ByteDance/Alibaba/Tencent, Ant Ling/Ming, decade-long memory contracts for Alibaba/Tencent/ByteDance, Gemini 4 Carbon's general-release date, NVIDIA Nemotron 3.5 Transcribe's official blog, distillation accusations (DeepSeek/Moonshot vs GPT-6 Astra), Meta's parental-reminder expansion, xAI's Mississippi home purchase in exchange for a dropped suit, Nvidia's reported Nscale guarantee, Grok Imagine Video 1.5 Lite, OpenAI's actual first-round raise, Senator Warren's reply from Anthropic (due Nov 3, no public reply visible by the end of the window), and Apple iOS 27.3 developer testing in China (Nov 4, due today).