Home

Frontline Hotspot

Fast-tracking AI industry hot events with concise ~1000-word analysis.

Kimi K3.1 Pro Weights Are Out: 2.8T Parameters, 1.56TB

At midnight Beijing time on November 5, 2026, Moonshot AI published the Kimi K3.1 Pro weights on Hugging Face, honoring the promise made when the model launched on October 23. The model has 2.8 trillion total parameters with about 104B activated per token, a 1048576-token context, 69 KDA plus 24 Gated DSA attention layers, a 401M vision encoder, and MXFP4 weights with MXFP8 activations under quantization-aware training, totaling roughly 1.56TB. The vendor-reported Terminal-Bench 4.0 score is 65.7; on the third-party Agent Arena of real user sessions it ranks first among open models with a net improvement rate of 5.29 percent, ahead of MiMo-V2.6-Pro at 2.68 percent (preliminary). The Kimi K3.1 Pro License requires a separate agreement for Model-as-a-Service businesses exceeding 20 million dollars in trailing 12-month revenue and prominent naming for products above 100 million monthly users or 20 million dollars monthly revenue.

Wuzhen Report Card: 4 Billion Downloads of Chinese Open AI

The 2026 Wuzhen Summit opened on November 3 with open source at center stage. Joe Tsai said on record that Qwen has 470+ open models, 4 billion cumulative downloads, and 330,000 derivative models; OpenRouter data shows DeepSeek V4.1-Flash topped global traffic at 33.6 trillion tokens in one week with the top five all Chinese models; a Cyberspace Administration report puts the top six called models all from Chinese teams. Zhejiang Lab 021 (236B), LingBot (2.2 million downloads in 9 months), and Tencent Hy4 preview were on display.

Tencent Ships Hy4, But the Weights Are Still Coming

On November 4, 2026 Tencent Hunyuan shipped the final Hy4 release, two months after open-sourcing the August 28 preview. It adds vision understanding and computer use for the first time and substantially scales pre- and post-training. Vendor-reported gains include GameDev +19.1 percent and GameLogic +9.8 percent, a WorkBuddy office score of 83.18 with 96.15 percent task pass rate, and 3.04 across 579 finance tasks judged by 213 experts. Hy4 is live on TokenHub, WorkBuddy/CodeBuddy, Yuanbao and ima, plus OpenRouter, at 6/18/0.3 yuan per million tokens for input/output/cache hits. Weights and a system card are promised soon but not yet released, and Tencent concedes a sizable gap to top closed models.

Kimi Writes Blender Scripts Now: 3D Modeling Gets Easy

Moonshot AI shipped 3D Make alongside K3.1 Pro: give it a description, a sketch, or a reference image, and the model infers shape, proportions, and spatial relations, then writes executable Blender, FreeCAD, or SolidWorks scripts and runs them in the cloud. Outputs are downloadable, previewable, and exportable to GLB, OBJ, STL and more, pitched as one-sentence 3D-printable models. Behind it sits the 2.8-trillion-parameter, 104B-active K3.1 Pro post-training push, with full weights promised before November 4.

Apple Intelligence Taps Qwen: China iPhone Gets AI Nov 4

On October 31, 2026, Alibaba and Apple announced a partnership to integrate the Qwen model into Apple Intelligence. Qwen opens in the latest iOS developer beta on November 4 and ships with iOS 27.3 in December for China users; it is reached through Siri, Writing Tools, and Visual Intelligence, with a first batch of a dozen-plus features including real-time translation and photo object removal. Users opt in, and Alibaba states requests are not used for training.

Doubao Makes Peanuts, Volcano MaaS Prints 4.8 Billion Yuan

Per LatePost's October 30 report, Volcano Engine's MaaS business took in 4.8 billion yuan in the first half with gross margin above 55 percent, and raised its original 15-billion-yuan annual target (new figure undisclosed), making commercialization the top priority. IDC's October 28 China AI software tracker must be read in two calibers: full-modality MaaS revenue hit 10.63 billion yuan with Volcano first at 45.1 percent (about 4.8 billion, corroborating LatePost) and Alibaba Cloud at 25.2 percent; but on the LLM-only caliber Alibaba Cloud leads at 46.6 percent, Zhipu 10.6, Volcano just 10.2 — reconciled by revenue mix: LLM 47.0 percent, video generation 32.7 percent (the Seedance base). Another figure, 49.5 percent, is Token-call share (June FORCE event), so three percentages live in three calibers. Context: Alibaba's AI Cloud segment posted 48.44 billion yuan for the quarter (+45 percent, segment caliber, not MaaS); Baidu government-enterprise bids totaled 1.385 billion yuan (project-based, not directly comparable); Doubao's consumer-side daily revenue of under 1 million yuan contrasts with MaaS's 4.8 billion; Seedance ARR of 2 billion US dollars at 70 percent gross margin is a point-in-time figure.

No.1 With 59, but You Can't Buy It: Gemini 4 Carbon's Gamble

Google shipped Gemini 4 Carbon on October 28, landing at 59 on Artificial Analysis' freshly rebuilt Intelligence Index v5.0 (private/held-out weight raised to 60 percent) — ahead of Opus 5.5 at 58 and GPT-6 Astra at 57, and 9 points over Argon's 50 from two weeks earlier; it takes first place in 4 of 9 evaluations (HLE-Diamond 69 percent, Terminal-Bench-Science 65 percent, AA-Omniscience, AutomationBench-AA), while the official table claims 15 of 17 leads, mostly self-reported. Three cybersecurity benchmarks — SEC-bench Pro 88.2 percent, ExploitGym 54.8 percent, CWE-bench 71.0 percent — are all Google self-reported, guardrail-free and unreproduced, with exploit validation the biggest jump (Argon 42.0). There is no API and no app entry; this week it opens only to vetted cyber defenders, expanding via the Fairwind Program. Introductory pricing matches Argon (2 dollars/10 dollars per million input/output tokens, 95 percent cached-input discount, doubling after) with no stated end date. AA's independent measurements add the other side: 7.76 dollars per index task, a 97.5-second median first-token delay, about 200 million output tokens per full run; on AA-Omniscience its wrong-answer conviction rate of 11 percent is above Argon's 4 percent yet far below Opus 5.5's and Astra's 50. The launch coincided with Alphabet earnings: down more than 10 percent after hours (multiple figures listed, none picked), and down 12.64 percent at the next close, the worst single day since the 2004 listing.

Google Genie 4 talks now - but only for two minutes?

Google DeepMind opened Genie 4 as a research preview on October 27 (trusted testers first, waitlist live in Project Genie): the first Genie-line world model to generate audio natively — ambient sound, music, and characters that talk back in real time, with interactive human speech called the true frontier. Gemini Omni grounds appearance and behavior while Gemini 3.8 Flash acts as the director that plans steps and checks frames, and in a robotics experiment Gemini 4 Argon decomposed a checkers task for the Apptronik Apollo humanoid to play out. Scene coherence of up to 2 minutes is the TechCrunch/The Verge reported figure (officially only drift limits are acknowledged), with generational comparisons anchored to the Project Genie prototype's roughly 60-second cap. Benchmarks, resolution, framerate, parameter count and pricing are all unpublished; every output carries SynthID. Runway's GWM Worlds 2 and the consolidation wave (AMD buying World Labs, Amazon in talks for Decart) close the piece.

GPT-6.1 Terra Cuts Agent Costs in Half: Here's the Math

OpenAI released GPT-6.1 Terra on October 26 Eastern Time (October 27 Beijing Time), built for execution and agent work. Official pricing says 50 percent below both the previous GPT-5.6 Terra (2/12 dollars per million tokens) and GPT-6.1 Sol — though 12 down to 5 dollars on output math works out to about 58 percent, and both figures are stated side by side. Official self-reported benchmarks: +15 percentage points on Agents' Last Exam at about one-third the cost, 50.3 percent on Terminal-Bench 4.0 versus 21.5, and factual error rate down from 53.6 to 16.7/4.3 percent; the independent Artificial Analysis index (48.1) is kept separate from official claims. Already wired into ChatGPT Work and Codex, with a seven-member family table (Luna/Terra/Sol/Ultrafast/Astra plus the two previous-generation models) sourced line by line.