Introducing GPT-6.1 Sol
OpenAI has released GPT-6.1 Sol, which it says comes close to Astra’s level on coding, computer use and professional work. Its API costs one-fifth of Astra’s standard price for input and output tokens.
Week of September 28, 2026
September 28, 2026 – October 4, 2026 · 36 storiesGoogle DeepMind announced Gemini 4 Argon, presented as a new era for its frontier models, while OpenAI recapped more than 20 DevDay announcements and released GPT-6.1 Sol at a fifth of Astra's API price. OpenAI also launched dots, proactive assistants for ongoing projects, and said it disrupted a coordinated campaign to distill its models' reasoning. On the infrastructure side, CoreWeave is bringing NVIDIA's next-generation hardware into production, and Amazon's $1 billion plan to address data center opposition drew criticism. In the U.S., a judge dismissed antitrust suits against Google's AI search, and authorities arrested a CEO accused of smuggling $300 million in Nvidia chips into China.
Model releases
Gemini 4 Argon: our next era of frontier intelligence
Google DeepMind has announced Gemini 4 Argon, which it presents as the start of a new era for its frontier models.
A model guide for the GPT-6 family
OpenAI has published a guide for startups on choosing between GPT-6 models. It covers tuning reasoning effort, writing prompts and skills, coordinating tools and preparing workflows for production.
Open source
Welcome RL Environments to the hub
Hugging Face now hosts reinforcement learning environments on its Hub.
Agents & harnesses
Are you a Codex Original?
OpenAI is collecting stories from developers, researchers and creators who use Codex for its Codex Originals program. Interested users can submit their project.
Holo4: powering generalist computer-use agents
A Hugging Face post presents Holo4, designed to power general-purpose agents that operate computers.
Introducing dots
OpenAI has launched dots, proactive assistants that keep working on complex projects and everyday tasks. The company says users stay in control while the work moves forward.
Getting the Source Right, Not Just the Fact: Source-Aware Verification for MCP Agents
A Hugging Face post describes a verification approach for MCP-based agents that checks where information comes from, not only whether it is correct.
AutoSynthData: Generating Training Data for Enterprise Agents
A Hugging Face post presents AutoSynthData, an approach for generating training data for enterprise AI agents.
Redefining enterprise intelligence with autonomous AI
An MIT Technology Review piece says enterprise AI is now fully operational, with model capabilities advancing faster than companies can adopt them. It cites a forecast that global AI investment will reach $2.5 trillion in 2026, up 44% on the previous year.
The Agent Said It Was Done. The Database Disagreed.
A Hugging Face post looks at cases where an AI agent reports a task as complete although the database shows otherwise.
Frontier & AGI
The eternal complement
An OpenAI essay argues that advanced AI may have its biggest impact on the routine work that supports breakthrough ideas. It looks at how execution could shape the economy and the pace of progress.
Research
Open TTS Leaderboard: Scalable Evaluation for Multilingual Text-to-Speech and Voice Cloning
Hugging Face has introduced an open leaderboard for evaluating multilingual text-to-speech and voice cloning models at scale.
With most information hidden, the game Stratego had stumped AI—until now
AI researchers have made progress on Stratego, a game where most information is hidden from players. The key was adding a second neural network that guesses what the opponent’s hidden pieces are.
Don’t be fooled—LLMs don’t reason
In an opinion piece for MIT Technology Review, a researcher who helped build the Go program that played the famous Move 37 in 2016 argues that large language models do not truly reason.
Products & apps
Basis completes a tax workbook 2x faster with GPT-6 Astra
Accounting AI firm Basis says GPT-6 Astra finished a 50-tab tax workbook twice as fast as GPT-5.6 Sol. Its better grasp of user intent also makes the company more confident using it in real conditions.
DevDay 2026 Recap
OpenAI has summarized more than 20 announcements from its DevDay 2026 developer conference. They cover GPT-6 Astra, ChatGPT, Codex, APIs, security and new developer tools.
The Den frees up 10-15 hours a week to grow with ChatGPT Work
Social club The Den uses ChatGPT Work as it opens a new location. Grant applications now take 2 hours instead of 3 days, and liquor-license paperwork 3 hours instead of 4 days.
How Albertsons Companies is reimagining retail from the inside out
U.S. grocery group Albertsons uses ChatGPT Enterprise and the OpenAI API. It aims to help its teams work faster and make shopping easier for millions of customers.
Chatham scales its capital markets expertise with OpenAI
Chatham Financial uses Codex and GPT-5.6 to build tools and redesign workflows. Trade validation now takes under 4 minutes instead of 30.
The latest AI news we announced in September 2026
Google has published a roundup of the AI updates it announced in September 2026.
Chips & compute
From Training to Production, NVIDIA and CoreWeave Close the Loop on Agentic AI
At its Fully Connected event in San Francisco, CoreWeave announced it is bringing NVIDIA’s next-generation infrastructure into production. The two companies have worked together on AI cloud infrastructure for nearly a decade.
Productive, Durable, Fungible: How NVIDIA AI Factories Maximize Return on Investment
NVIDIA estimates that each megawatt of AI data center capacity costs about $60 million to build. The company sets out the factors it says determine the return on such investments, starting with earning capacity.
How NVIDIA GPUs Help Accelerate OpenAI’s GPT-6 Astra Ultrafast
GPT-6 Astra Ultrafast, which runs on NVIDIA Blackwell GPUs, is now available in the OpenAI API and to eligible ChatGPT Work and Codex users. It generates tokens up to 8 times faster than Astra’s standard mode.
NVIDIA DGX Spark 64GB Gives Developers More Ways to Build and Scale Local AI
NVIDIA will release a version of its DGX Spark desktop AI computer with 64GB of unified memory this month, built by partners including Acer. The company links it to open models becoming small enough to run locally.
The 7-year-old Nvidia Shield TV is now $100 more expensive due to AI
The Nvidia Shield TV Pro, first released in 2019, now sells for $299.99, a $100 increase that Ars Technica attributes to AI.
US arrests tech CEO accused of smuggling $300M in Nvidia chips into China
U.S. authorities have arrested a tech company CEO accused of smuggling $300 million worth of Nvidia chips into China. The case is the latest in a series of arrests over illicit exports of Nvidia hardware.
Amazon’s $1B plan to combat data center backlash draws more backlash
Amazon has announced a $1 billion plan to address opposition to its data centers. It was praised for ending non-disclosure agreements but criticized for downplaying data center pollution.
Business & funding
The Lenfest Institute grows landmark program with expanded OpenAI support
OpenAI is expanding its support for the Lenfest AI Collaborative and Fellowship Program, which works with news organizations. It is providing $5 million in funding plus up to $5 million in software credits and engineering help.
Helping small businesses put AI to work
OpenAI is partnering with America’s SBDC network to offer small businesses hands-on AI training and local support. It also released a report on how small teams use AI.
Policy & regulation
Judge dismisses Chegg and Penske antitrust lawsuits targeting Google AI search
A U.S. judge has dismissed antitrust suits brought by Chegg and Penske over Google’s AI-powered search. The court recognized that AI search has consequences but ruled it is not an antitrust matter.
Safety & security
How we will do better for Australia
OpenAI has apologized for incidents involving Australian government websites. It describes stronger safeguards and further support for the country’s cyber defenses.
Towards safety cases for frontier AI training
OpenAI has published early guidelines for safety cases covering the training of frontier models. They address technical safeguards, operational practices and how to investigate misalignment incidents.
Disrupting a coordinated model-distillation campaign
OpenAI says it disrupted a campaign that tried to extract its models’ protected reasoning. The company is reinforcing its defenses against this kind of adversarial distillation.
Introducing SynthID Bio
Google DeepMind has presented SynthID Bio, a proof of concept for watermarking proteins designed by AI. The watermark is meant to leave the proteins’ biological function intact.
Apple changes full-disk access permissions to curb abuse from AI agents
Apple has changed how full-disk access permissions work on its systems to limit misuse by AI agents. Meta argues the permission should not be enough to let its Muse agent read messages, while Apple disagrees.