Two weeks ago Google announced its best model ever and then wouldn’t let you use it.

Yesterday that changed.

Gemini 4 Argon is now live in the Gemini app, AI Mode, the API and Antigravity for anyone on Google AI Pro ($20 a month) or Ultra.

Free and Plus users are still locked out. 🙄

Here’s why Google sat on it since September 30.

Argon was trained to hunt and patch software vulnerabilities on its own.

So Google watches its chain of thought mid-task and kills the run if it wanders.

Translation: the emergency brake Nadella asked for, shipped with the model attached. 🫡

Now the numbers that matter to YOUR wallet.

API pricing is $2 per million input tokens and $10 per million output.

On December 1 both numbers double. ⏰

One wrinkle before you cancel your other subscriptions.

Bloomberg reported on launch day that Google’s own employees think Argon crushes benchmarks but stumbles on real coding, especially front-end design.

Google called that inaccurate.

Its own table shows Argon beating Claude Opus 5.5 on long software tasks and losing to it on command-line coding, 57.4% to 66.4%.

Artificial Analysis scores it 53, level with GPT-6 Astra and Claude Fable 5.1, 5 points behind Opus 5.5.

So here’s the play.

Argon is cheap for 7 weeks.

Run it against whatever you’re using now on YOUR actual work, not a leaderboard, and decide before the bill doubles.

The labs are now shipping models with a kill switch bolted on and a sale price with an expiry date.

Somebody has to know which one is worth paying for, and that somebody gets paid. 💰

Argon is the gas inside most incandescent light bulbs.

It’s there because argon is too lazy to react with anything.

The filament glows without burning out.

A whole model named after the element whose entire job is not doing anything it isn’t supposed to. 😂

Anthropic opens its books today.

CNBC reports the company is walking a select group of institutional investors through its Q3 financials at its San Francisco headquarters this afternoon.

It’s the first time anyone outside the building has seen those numbers.

Bloomberg first reported the meeting.

This is the final rehearsal before a formal IPO roadshow that could start the week of November 9, with a listing targeted before Thanksgiving.

Investors are reportedly modelling a valuation between $1.8 trillion and $2 trillion.

For context, Anthropic’s last private round in May reportedly valued it at roughly $965 billion.

So the ask is “we doubled in five months, pay up.” 🤯

Why the numbers matter beyond Anthropic.

Private AI revenue figures have been moving public markets all year.

Chip stocks sold off this week on exactly that nervousness.

If Q3 lands big, the whole AI trade gets a tailwind.

If it disappoints, the IPO window gets a lot smaller for everyone behind Anthropic in the queue.

Now sit with the week this is landing in.

Monday I told you Anthropic cut its own models off the internet after one filed a fake police tip.

Yesterday I told you its red team said small models could be copying themselves onto TV boxes by Christmas.

Today it asks for $2 trillion.

Anthropic hasn’t confirmed any of the IPO dates and declined to comment on the meeting.

What a time to be alive. 🙌

Go switch your model picker to Argon and run one real task from your business through it today.

If it beats what you’re paying for, you’ve got until December 1 to lock in the cheap seats. 🔥

Jeff J. Hunter

Founder, AI Persona Method | TheTip.ai

Get paid to solve problems like this → AI Certified Consultant

PS. Google scrapped Gemini 3.5 Pro, promised for June, to build Argon instead. Bloomberg Intelligence puts a single frontier training run at up to $400 million. Argon isn’t just a model launch, it’s Google’s apology for the one that never shipped. 👀

“AI Money Group” to Learn to Make Money with AI