AI Weekly · Week 40, 2026 · 09.28 — 10.04
GPT-6.1 Sol and Gemini 4 Argon arrive in the same week
AI summary
This issue covers 28 September to 4 October 2026. Model releases came thick and fast: at DevDay OpenAI introduced GPT-6.1 Sol, saying its agentic coding, computer use and professional work abilities approach GPT-6 Astra at one-fifth the price; Google DeepMind released the frontier model Gemini 4 Argon, first to trusted cyber defenders, plus SynthID Bio, which embeds verifiable watermarks in AI-generated proteins; and AI2 released Olmo-core 3 for trillion-parameter MoE and open-sourced AstaBrief 8B. On the industry side, AMD agreed to buy World Labs in an all-stock deal worth about $8.2bn, with Fei-Fei Li to become executive vice-president and chief scientist; the FTC opened a consumer protection investigation into OpenAI, Anthropic and others; and OpenAI said it had blocked a model-reasoning theft attack, though the technique still works on Azure. On infrastructure, NVIDIA and CoreWeave put Vera Rubin NVL72 into production and DGX Spark gained a 64GB version.
Two frontier models race ahead
In shortAt DevDay OpenAI released GPT-6.1 Sol, saying it approaches GPT-6 Astra's intelligence on agentic coding, computer use and professional work while costing one-fifth as much for standard input and output tokens; GPT-6 Astra Ultrafast already runs on NVIDIA Blackwell GPUs and is available through the API and to eligible ChatGPT Work and Codex users. Google DeepMind released the frontier model Gemini 4 Argon in the same period, first to trusted cyber defenders in the Fairwind Program and later to developers, enterprises and consumers, starting at $2 per million input tokens and $10 per million output tokens.
OpenAI launches GPT-6.1 Sol, says it nearly matches GPT-6 Astra and costs less
OpenAI says GPT-6.1 Sol delivers significant improvements over GPT-6 Sol across complex professional tasks, including code writing and debugging, document understanding, and executing multistep business workflows.
Gemini 4 Argon: our next era of frontier intelligence
Gemini 4 Argon delivers frontier performance in complex workflows across real-world software engineering, enterprise knowledge work like legal and finance, and cybersecurity defense. Today, we’re announcing our new frontier model, Gemini 4 Argon, which is rolling out to a set of trusted cyber defenders through our Fairwind Program.
How NVIDIA GPUs Help Accelerate OpenAI’s GPT-6 Astra Ultrafast
GPT-6 Astra Ultrafast, running on NVIDIA Blackwell GPUs, is available now in the OpenAI API and to eligible ChatGPT Work and Codex users. Accelerated by inference optimizations through OpenAI’s models that tap into the capabilities of the NVIDIA Blackwell architecture, Ultrafast offers up to 8x faster token generation than the Astra Standard mode.
AMD acquires World Labs
In shortAMD announced it will acquire World Labs, a world-model startup focused on spatial intelligence, in an all-stock deal worth about $8.2bn. The deal is expected to close by the end of 2026, subject to regulatory approval. World Labs founder Fei-Fei Li will join AMD as executive vice-president and chief scientist, reporting directly to CEO Lisa Su and leading frontier research. It is the largest industry consolidation of the period and shows chipmakers accelerating their push into world models and spatial intelligence.
AMD buys AI world model startup World Labs for $8.2 billion
Fei-Fei Li's AI startup World Labs is being acquired by AMD for $8.2 billion. Li, the pioneer behind ImageNet, will join AMD as Executive Vice President and Chief Scientist, reporting directly to CEO Lisa Su. World Labs brings its spatial intelligence models, including Atlas, which generates and simulates 3D environments from minimal input.
Regulatory and safety pressure mounts
In shortThe US Federal Trade Commission is investigating OpenAI, Anthropic and other leading AI labs over potential consumer protection violations. Chair Andrew Ferguson plans to compel document production and question executives through legally binding civil investigative demands, expected within weeks. Meanwhile OpenAI said it had blocked a distillation campaign that stole model reasoning content. The activity appeared at small scale from 1 July, then surged on 24-25 July to 16,000 requests from more than 4,000 users across over 15,000 linked accounts. OpenAI said it shut it all down on 28 July, but the technique still works on Azure.
FTC launches sweeping probe into OpenAI, Anthropic, and other AI labs over consumer protection concerns
The FTC is formally investigating OpenAI, Anthropic, and other leading AI labs over potential consumer protection violations. The agency plans to force document handovers and executive testimony through legally binding demands. The probe was already underway before the Hugging Face hack and goes well beyond the FTC's earlier push to hold companies liable for their agents' behavior.
OpenAI says it stopped a campaign to steal its models' reasoning, but the trick still worked on Azure
OpenAI says it stopped a coordinated campaign in which more than 15,000 accounts tried to copy the hidden reasoning of its models. The company ties part of the activity to people connected to Moonshot AI. But researchers found that the same attack kept working on Microsoft Azure for weeks, even against the new GPT-6 Astra.
Agent evaluation and training infrastructure
In shortThe Microsoft Copilot Studio team and Hugging Face jointly released ThinkingBox, which runs agents in isolated MCP tool sessions and then performs executable checks on final backend state and side effects, rather than only judging answer text or whether tool calls were valid, offering a new paradigm for agent evaluation. AI2 released Olmo-core 3, a redesigned open MoE training system aimed at scaling MoE training to trillion-parameter size while keeping compute efficiency; it also open-sourced AstaBrief 8B, which turns research questions and retrieved literature snippets into cited reports, with weights and training data released too.
The Agent Said It Was Done. The Database Disagreed.
Microsoft ThinkingBox grades AI agents on the records they leave behind, not the sentences they generate, and then asks whether they can do it twenty times in a row. It is now available through Hugging Face.
Introducing Olmo-core 3: Open, scalable training infrastructure for large MoEs
Today we’re releasing Olmo-core 3, a significant upgrade to our framework for developing large language models featuring a redesigned open mixture-of-experts (MoE) training system. Olmo-core 3 is designed to scale MoE training into the trillion-parameter range while preserving computational efficiency.
Open-sourcing AstaBrief, the fast report-generation model in Asta
We built AstaBrief 8B, a model that turns a research question and retrieved literature excerpts into a cited report. AstaBrief is available in Asta's Generate a report feature today as Fast mode alongside Claude-powered Thinking mode, and we're also open-sourcing it and the training data so other
Compute and privacy tech reach production
In shortCoreWeave announced that NVIDIA Vera Rubin NVL72 systems with Spectrum-X 102.4T Ethernet are available on CoreWeave Cloud, with Cognition the first customer running the system in production. NVIDIA will launch a 64GB unified-memory version of DGX Spark this month through Acer, ASUS, Dell, Gigabyte, HP and MSI, on sale 23 October from $4,999. Google Research released a next-generation federated learning system based on trusted execution environments, offering verifiable, auditable guarantees for data anonymisation; Gboard has adopted it to launch English and Japanese next-word prediction models.
From Training to Production, NVIDIA and CoreWeave Close the Loop on Agentic AI
Building on nearly a decade of co-engineering, CoreWeave has built NVIDIA compute, networking and software into a cloud purpose-built for AI that’s still returning on investment across multiple generations of deployment. Now, CoreWeave is bringing the next generation of NVIDIA infrastructure to production.
NVIDIA DGX Spark 64GB Gives Developers More Ways to Build and Scale Local AI
Local AI is becoming more useful by the token. As AI agents move from experiments into everyday development, increasingly capable open models are shrinking to fit on more devices, giving builders more to run locally.
Toward provably private learning from federated data
In “Toward provably private learning from federated data”, we announce the next generation of our FL system, which leverages Trusted Execution Environments (TEEs) to provide verifiable and auditable guarantees for data anonymization. Gboard has adopted the system to launch next-word prediction models in English and Japanese, with stronger privacy guarantees and improved accuracy.