This Week in AI Tools: GPT-6.1 Sol, Gemini 4 Argon, Runway Ads

This week in AI tools, four stories mattered: a new mid-priced OpenAI model, a Google frontier model you can't have yet, an ad product from Runway that writes, publishes and measures campaigns, and a court ruling that AI companies will be reading closely. Here's what each one means for your workflow.

OpenAI's GPT-6.1 Sol: near-Astra quality at a fifth of the price

On 30 September, OpenAI introduced GPT-6.1 Sol, an upgrade to GPT-6 Sol that it says "nearly matches GPT-6 Astra's intelligence on agentic coding, computer use, and professional work at one-fifth of Astra's standard input and output token prices." For developers the headline is the price: $2 per million input tokens and $10 per million output, with cached input at $0.10.

OpenAI's own benchmarks claim the model matches Astra on a real-codebase software engineering test and reduces factual errors on hard prompts from 11.4% to 7.7% at low reasoning effort. The company also says it's more likely to admit when a tool is broken instead of guessing.

What it means for ChatGPT users: the post says GPT-6.1 Sol is available to Plus, Pro, Business, Enterprise and Edu users in ChatGPT Work and Codex, but "not yet available in Chat." Free users and ordinary chat sessions don't get it yet. If you pay for an API-based tool, expect its vendor to switch models and, with luck, pass on the saving.

Google's Gemini 4 Argon goes to cyber defenders first

On 30 September, Google announced Gemini 4 Argon, which it describes as delivering "frontier performance in complex workflows across real-world software engineering, enterprise knowledge work like legal and finance, and cybersecurity defense." The model's output limit jumps to 1 million tokens, from 64,000 before, which Google says suits long multi-step tasks.

The unusual part is the rollout. Argon is going first to "a set of trusted cyber defenders" through Google's Fairwind Program, because, in Google's words, "Argon can autonomously find, validate, and patch critical software vulnerabilities." Google says it wants it with developers, enterprises and consumers "as soon as possible," starting with paid API customers and Google AI Ultra subscribers. For now, the Gemini you use is unchanged.

Runway Ads: generate, publish, measure, repeat

Also on 30 September, Runway launched Runway Ads, an engine that "generates high-quality, on-brand video and image ads," publishes approved variants to Meta, Google and TikTok, "reads performance back from each platform and regenerates the next round around what's earning spend." Human approval is on by default; teams can hand over publishing campaign by campaign.

Runway says it built the product on its own marketing: since July it scaled from 77 to roughly 900 ads a week while doubling return on ad spend. Treat those as a vendor's numbers about itself. The product is in a pilot with "select enterprise partners," with no public pricing, so small advertisers are watching rather than buying. Our Runway review covers the video tools you can use today, and the best AI video tools list has alternatives.

An appeals court rejects fair use for AI training

On 30 September, the US Court of Appeals for the Third Circuit sided with Thomson Reuters against Ross Intelligence, rejecting Ross's argument that training a legal search tool on Westlaw's headnotes was fair use. Claims Journal called it "a first-of-its-kind ruling by a US appeals court in a wave of high-stakes copyright cases over AI training."

Two caveats before anyone declares a winner. The court's reasoning is sealed for now, and the case "does not involve generative AI," as the report notes, so it may not transfer directly to lawsuits against chatbot and music-generator companies. Still, it's the first appellate answer on the question, and it went to the copyright holder. For how this plays out in music, where Suno faces a new lawsuit from Universal and Sony, see our Suno vs Udio comparison; a full guide to AI music licensing follows next week.

Also this week in AI tools

• ElevenLabs for Students. ElevenLabs launched a student plan on 25 September: three months of its creative, agents and API products plus a year of its ElevenReader Ultra app, for university students aged 18 and over in the US, Canada, the EU, Australia and the UK, "with more countries to follow." India isn't included yet. See our ElevenLabs review.

What to do about it

If you build on the OpenAI API, test GPT-6.1 Sol against your current model on your own prompts before switching; the price cut is real, but the post itself says Astra remains better for the hardest tasks. Everyone else can wait: Argon isn't available, Runway Ads is a pilot, and the court ruling changes nothing about the tools you use today.

FAQ

Is GPT-6.1 Sol available in ChatGPT?

Not in ordinary chat yet. OpenAI says it is available to Plus, Pro, Business, Enterprise and Edu users in ChatGPT Work and Codex, and to developers through the API at $2 per million input tokens.

Can I use Gemini 4 Argon?

Not yet. Google is rolling it out first to trusted cyber defenders through its Fairwind Program, then to paid API customers and Google AI Ultra subscribers, with no public date for the Gemini app.

Does the Thomson Reuters ruling affect ChatGPT or Suno?

Not directly. The Third Circuit case concerned a legal search tool, not generative AI, and the opinion is sealed. It is the first appellate ruling against fair use for AI training, so later cases will cite it.