TL;DR: Two weeks after a limited debut, Google has made Gemini 4 Argon, its new frontier AI model, available worldwide. Developers can use it through the Gemini API and AI Studio, businesses through Gemini for business and Google Cloud, and consumers in the Gemini app and AI Mode in Search if they pay for Google AI Pro or Ultra. It launches at an introductory $2 per million input tokens and $10 per million output tokens until November 30, with a 1 million token limit. Google’s own benchmarks are strong, but early hands-on coding reviews from developers are decidedly mixed.
From cyber defenders to everyone
When Google first announced Argon on September 30, it went only to a small group of trusted cybersecurity defenders through its Fairwind Program. As we noted at the time, the flagship wasn’t for regular users yet.
That changed on October 13. In a post on Google’s blog, Koray Kavukcuoglu, SVP at Google DeepMind and Google’s Chief AI Architect, said Argon is “now globally available” across Google’s main AI products. Google describes it as a model for “deep work in complex, real-world tasks” in coding, finance, legal, cybersecurity and science.
Where you can use Gemini 4 Argon
| Who | Where | What you need |
|---|---|---|
| Developers | Gemini API, Google AI Studio, Android Studio | Paid API tier; AI Ultra subscribers can also try it in AI Studio |
| Agentic coding | Google Antigravity | Google AI Ultra subscription (or business access) |
| Enterprises | Gemini for business, Gemini platform API, Antigravity | Google Cloud / Gemini for business account |
| Everyday users | Gemini app and AI Mode in Search | Google AI Pro or Ultra subscription |
| Scientists | Antigravity’s highest usage tier, free for six months | Application; 10,000 places, subject to selection |
| Security teams | Fairwind Program | Application; Google says it is expanding to hundreds more partners |
On Google Cloud, enterprise developers can call the model under the ID gemini-4-argon, according to the Google Cloud blog. Free users of the Gemini app don’t get Argon.
Pricing and the 1 million token limit
Google is pricing Argon aggressively for launch:
- Introductory price (until November 30, 2026): $2 per million input tokens and $10 per million output tokens.
- Cached input: 95% off the input price.
- Standard price after November 30: $4 per million input tokens and $20 per million output tokens.
The model also comes with what Google calls an “industry-leading” 1 million token unified input-output limit through the Gemini API. That means the prompt and the answer share one 1M budget, and Google points developers to a new “continuation protocol” for long jobs. The model card says Argon accepts text, images, audio and video, is a sparse mixture-of-experts model trained on Google’s TPUs, and has a knowledge cutoff of March 2026.
What Google says Argon can do
Google filled its launch post with internal results. These are the company’s own claims and haven’t been independently verified:
- Inside Google, Argon in Antigravity helped teams increase agentic code submissions by more than 35% while reducing code rollbacks by nearly a third.
- It helped the Chrome team split core webpage handling out of a roughly 6 million line legacy folder, finishing “months of work in weeks.”
- It sped up Google’s open Gemma models by 2.85 times end to end on an A100 GPU through vLLM, with no quality drop, Google says.
- It made the open source quantum chemistry package xTB up to 2.3 times faster at predicting molecular structures.
Science gets a big push too. Google says Argon nearly doubled the inference throughput of its AlphaQubit quantum error-correction decoder, and helped researchers find a broader class of counterexamples to “Grad’s conjecture”, a decades-old assumption in fusion research. To encourage more of this, Google is offering 10,000 scientists free access to Antigravity’s highest usage tier for six months.
How it stacks up on benchmarks
Google says Argon ranks first on the Vals Index, which measures performance across finance, coding, legal and tax work. Its model card compares Argon with OpenAI’s and Anthropic’s latest models, and the picture is more nuanced than “best at everything”:
| Benchmark | Gemini 4 Argon | GPT-6 Astra | Claude Fable 5.1 | Claude Opus 5.5 |
|---|---|---|---|---|
| Vals Index (knowledge work) | 68.9% | 63.1% | 65.8% | 67.0% |
| DeepSWE v1.1 (agentic coding) | 77.9% | 74.1% | 67.4% | 74.2% |
| FrontierSWE v2 (agentic coding) | 55.0% | 65.5% | 56.3% | 62.3% |
| Terminal-bench 4.0 | 57.4% | 58.2% | 57.9% | 66.4% |
| GraphWalks, 256k to 1M tokens | 84.2% | 71.8% | 65.0% | 66.8% |
Argon clearly leads on long-context and knowledge-work tests, but trails rivals on some of the toughest coding benchmarks, such as FrontierSWE and Terminal-bench.
Developers’ early verdict: impressive, but uneven at coding
That benchmark split shows up in early hands-on testing. On a busy Hacker News thread, a benchmarking group called Gert Labs said it tested Argon in 100 coding environments and found it roughly between Anthropic’s Fable 5 and Fable 5.1, “maybe 2 months behind the frontier.” The team also said it caught Argon in infinite reasoning loops, which it called the first time it had seen that in a near-frontier model “in a long time.”
Developer and YouTuber Theo Browne was harsher. In a long hands-on review, summarized by Post-Cutoff, he praised Argon’s multimodal and front-end skills but said he had not merged any code the model produced during his real-world testing. Other commenters were far more positive about its writing, saying it answers more directly and with clearer structure than rivals, which matches Google’s own claim that Argon “skips clichés” and hedges less.
Google’s model card is also candid about limits: Argon can hallucinate, may show “occasional slowness or timeout issues”, and Google says it could not rule out that the model reached its internal danger thresholds for cyber and biology capabilities, so it shipped extra safeguards.
Why it matters
Argon is Google’s strongest answer yet to OpenAI’s GPT-6 family and Anthropic’s Claude 5 models, and the introductory pricing undercuts many rivals at the frontier tier. For businesses that work with huge documents, contracts or data sets, the 1M token limit and strong long-context scores are the headline. For coders, it’s worth testing, but early reports suggest keeping your current tool close by. If you want to understand the “agentic” work Google keeps talking about, our explainer on what agentic AI is covers the basics.
The bottom line
Gemini 4 Argon is now available to paying users and developers almost everywhere Google offers AI, at a launch price that doubles after November 30. Google’s numbers paint it as a frontier leader in knowledge work and long-context tasks. Developers are less sure about its coding reliability. Try it on your own real tasks before you switch.
The Teck Hustlers Editorial Team covers technology, business, finance and gaming news for readers who want the story behind the headline. Every article is researched, written and edited by our in-house team, checked against company announcements, filings and reputable reporting, and updated when new facts emerge. We focus on what each development means for founders, professionals, investors and everyday users. Have a tip or a correction? Reach us through our contact page.
