// 9TO5GOOGLE — MOBILE & WEB
Gemini 4 Argon must reverse Google’s AI inertia
Good morning and happy October. Today, I’m cringing at this gaggle of billionaires pretending to be human, vibing to the new Tinashe album, Popstar, and heartily dining on all the brutal ‘Digger’ reviews.
As a reminder, Inbox is the tech newsletter that goes beyond the news cycle. We cover the most important 9to5Google stories from the last couple of days along with a few highlights from around the web. We publish every Tuesday and Thursday, and if you like what you read, please subscribe.
Inbox is written by Daniel Bader. Read previous issues here, and then catch up with Ben Schoon’s excellent Weekender column published every Sunday.
After months of waiting, Gemini 4 ‘Argon’ is finally out — sort of. Google unveiled the new state-of-the-art model late yesterday but has only made it available to “a set of trusted cyber defenders” while it undergoes testing. The company says it will get a wider release “as soon as possible.”
This is a big moment for Google. Many users and critics alike wrote it off completely, with one analysis firm saying “for all intents and purposes, we believe DeepMind is no longer a frontier lab.” While the company has released several useful, fast and affordable Flash-branded models in 2026, its latest salvo into state-of-the-art AI was back in November 2025 with Gemini 3 Pro. In fact, it was so good that for a brief time Google was considered to be well ahead of OpenAI and Anthropic, and presaged a “code red” inside the frontier labs.
It’s hard to overstate how important it is for Gemini 4 to succeed. After a massive executive shuffle in August saw Demis Hassabis and Jeff Dean removed from the core management team, most of DeepMind’s operations were shifted from London to Mountain View, where new head Koray Kavukcuoglu reports directly to Google CEO Sundar Pichai.
While there has been speculation that Google, with its entrenched customer base, would walk away from the frontier in favor of cheaper, more product-focused pursuits, Kavukcuoglu has refuted those rumors. Earlier this month, he said, “To put it very bluntly, there’s nothing other than being at the frontier that is important for us. We have the whole full stack… and we are able to optimize to be able to achieve our goals.”
According to benchmarks, Gemini 4 is competitive with GPT-6 Astra and Claude Opus 5.5 in most tests, and is trained to be much better at agentic tasks than any previous Gemini model. It’s also better aligned, with a focus on cybersecurity, and has a massive 1M token output limit, which makes it far more capable of producing large bodies of work in one prompt with less compaction.
The big question, though, is whether Argon will be competitive in the real world. A Bloomberg report published shortly before the embargo lifted on the Gemini 4 news quoted anonymous Google employees skeptical of its coding abilities. From Bloomberg: “While Gemini 4 has performed well on benchmarks widely used to gauge model efficacy, it does less well when employees actually put it to work, according to people with direct access to the effort. The model struggles to handle certain coding tasks, said the people, who requested anonymity to discuss an internal matter.”
Argon is also curiously priced. Its introductory rate of $2/$10 per million input/output tokens matches OpenAI’s Sol 6.1 cost right now, but will eventually double to hit the $4/$20 price of Claude Opus 5.5. While still significantly cheaper than Astra or Fable, it has to prove its mettle with users before Google is welcomed back into the arms of the fickle AI masses.