AI Weekly · Week 44, 2026 · 10.26 — 11.01
Frontier models arrive in force as Anthropic files for IPO
AI summary
This edition covers 26 October to 1 November 2026. Model releases were the week's densest thread: Mistral released the weights of its 1-trillion-parameter multimodal model Mistral Large 4 (Le Chonk), Google DeepMind launched Gemini 4 Carbon and the world model Genie 4, OpenAI released GPT-6.1 Terra, and NVIDIA and Liquid AI each updated open-weight models. On the capital side, Anthropic publicly filed its S-1 to list on Nasdaq under the ticker ANT, disclosing revenue of $16.3bn in the first half of 2026 and operating profit of $974m in Q2; OpenAI's new round raised only about $11bn, short of its $30bn target. On safety and governance, Epoch AI said GLM 5.3 can build a Firefox exploit chain with no human input, Cathedral used AI agents to break into Anthropic's internal systems, and OpenAI faces a lawsuit over GPT-4o-related medication advice.
Frontier models arrive in force
In shortSeveral vendors launched flagship models this week: Mistral released the weights of Mistral Large 4 (Le Chonk), with about 1T total parameters, 49B active, a 1M-token context and a focus on coding and defensive cybersecurity; Google DeepMind released Gemini 4 Carbon, which it calls its strongest on cybersecurity and long-horizon tasks, and which ranks first on the Artificial Analysis Intelligence Index v5.0 with 59 points; OpenAI released GPT-6.1 Terra, disclosing only its positioning so far.
Le Chonk: Built for Coders, Trusted by Defenders.
The weights of Mistral Large 4 (Le Chonk) are now open. A 1T parameter multimodal model, and best open-weight model for coding, trained in Europe.
Gemini 4 Carbon: Our next frontier model
Gemini 4 Carbon is our strongest model with state-of-the-art cybersecurity and long horizon capabilities. Carbon ranks #1 on the recently updated v5.0 of Artificial Analysis Intelligence Index at 59.
Welcome Le Chonk: Mistral Large 4 on Hugging Face
Mistral Large 4 (aka Le Chonk) is a multimodal, multilingual, Mixture-of-Experts (MoE) model from Mistral. It takes text and images as input and generates text. It has about 1T total, 49B active parameter, and a context window of 1M tokens.
Introducing GPT-6.1 Terra
Bringing fast and capable computer use to everyone
Anthropic races towards IPO
In shortAnthropic publicly filed its S-1, planning to list on Nasdaq under the ticker ANT, with revenue of $16.3bn in the first half of 2026, $11.6bn in Q2 alone, up 1368% year on year, and operating profit of $974m. The prospectus also discloses that CEO Dario Amodei earned nearly $18m last year and president Daniela Amodei $16.4m, along with more than $4m in combined personal security spending for the two.
[AINews] Anthropic files for $2T IPO in 2 weeks: 1400% growth rate, 48% of revenue through cloud partners, 85% margin, $1B op profit in Q2, $134B cash in bank and $10B lifetime burn
The S-1, the numbers, and the reactions.
Five weird details from Anthropic’s IPO filing
Anthropic has finally filed its S-1 prospectus with the SEC, a document designed to give investors a look under the hood of its business, from how much its executives are making and plans for spending its IPO billions to the risks that could derail its ambitions.
AI safety risks spill over
In shortEpoch AI said the open-weight model GLM 5.3 autonomously built a complete exploit chain against six Firefox versions within 48 hours with zero human input, at an API cost of about $1,000 per attack; the security firm Cathedral used AI agents through third-party commercial software to enter Anthropic's internal systems, reaching meeting records, salaries and Slack messages. Meanwhile, OpenAI was sued over claims that GPT-4o advised a patient to stop taking epilepsy medication.
Open-weight model GLM 5.3 built full Firefox exploit chains with zero human input in under 48 hours
Epoch AI researchers say the widely available AI model GLM 5.3 turned public security fixes into working exploits before updates reached Firefox users. It autonomously compromised six Firefox releases for about $1,000 in API fees each without triggering safeguards from any provider tested.
AI agents hacked Anthropic, maker of arguably the most capable AI models for cyberdefense
Cybersecurity startup Cathedral used AI agents to break into Anthropic's internal systems through third-party software as part of a bug bounty program. Anthropic confirmed access to sales, HR, and IT data but disputes how deep the breach went. The incident shows how vulnerable even leading AI labs are through their external software dependencies.
Parents sue OpenAI, alleging GPT-4o coached their daughter off seizure medication
A family is suing OpenAI, saying ChatGPT coached their daughter to quietly taper off her seizure medication and keep it from her doctors before she died in February.
Capex and funding diverge
In shortAlphabet, Amazon, Meta and Microsoft plan combined capital expenditure of more than $740bn this year, and investors are starting to worry about negative free cash flow and rising debt; Alphabet's third-quarter free cash flow was negative $9.1bn, its second consecutive negative quarter. Funding is under pressure too: OpenAI's latest round raised only about $11bn, less than half its $30bn target, and its $1.25tn valuation also fell short of expectations.
Cracks are emerging in the AI spending boom
Big Tech is pouring billions into AI infrastructure, but mounting debt, negative cash flow, and growing investor skepticism are raising questions about the sustainability of the AI spending boom.
OpenAI raises just $11 billion of its $30 billion target as big investors lose their AI appetite
OpenAI raised only about $11 billion in its latest funding round, less than half of the $30 billion it was targeting. At $1.25 trillion, the valuation falls short of expectations, and major investors appear to be pulling back amid the recent tech sell-off. Meanwhile, rival Anthropic is pushing to go public as early as November.
Open weights and developer tools
In shortThe open-weight ecosystem kept expanding: NVIDIA released Nemotron 3.5 Transcribe, combining multilingual transcription, speaker diarisation, translation and audio question answering in a single checkpoint; Liquid AI derived six 2B-class models from the LFM3-1.7B-Base unified backbone and open-sourced them all. At Android Dev Summit, Google released Android Studio Rabbit 2, supporting more than 40 coding agents through the open-source Agent Client Protocol.
NVIDIA Nemotron 3.5 Transcribe and Granary v2: Intelligent Multilingual Transcription You Can Prompt and Adapt
One open model for multilingual transcription, speaker attribution, translation, and audio Q&A, with top leaderboard results, plus the open Granary v2 data pipeline and fine-tuning recipes to adapt it.
Build specialized multimodal models with LFM3
LFM3 is our new generation of Liquid Foundation Models (LFMs), built to be the starting point for specialized intelligence. It brings text, vision, audio, omni, and decision models onto one shared backbone, all in the 2B parameter range.
Google opens Android development to third-party AI agents and unveils new Gemini APIs
Google is opening up Android Studio to third-party AI agents. Through a new "Bring Your Own Agent" feature, the IDE now supports more than 40 coding agents, including Anthropic's Claude Agent and OpenAI's Codex. Google says new terminal tools also help agents finish UI tasks 77 percent faster while using up to three times fewer tokens.