Google is releasing Gemini 4 Carbon to the general public, two weeks after a limited release to a small group of “cybersecurity defenders.” According to The Verge AI, Google calls Carbon its “strongest model” yet and says it can compete with the top systems from Anthropic and OpenAI. Google is also previewing a “swarm” feature in its Antigravity agent platform, which turns a single prompt “into a coordinated team of agents.”
The Quick Version
- What: Gemini 4 Carbon, Google’s new top-tier model, aimed at the hardest reasoning and coding work.
- Price: $2 per million input tokens and $10 per million output tokens during an introductory period. That’s around a fifth of what OpenAI charges for GPT-6 Astra and a quarter of what Anthropic charges for Fable 5.2.
- Catch: Prices double at the end of November. Even then, Carbon stays well below its rivals per token.
- Who gets it: Paying developers and enterprises through the Gemini API, Google AI Studio and Gemini for business. Google AI Ultra subscribers get it in the Gemini app, AI Mode in Search and Antigravity.
- Who doesn’t (yet): Free users and lower-tier subscribers. Google hasn’t said when, or if, they’ll get access.
Carbon vs. Argon: Which One Do You Need?
Google now has two frontier models, and it’s being unusually clear about how they differ. Argon launched only last month, and Google says it will likely remain “the ideal choice for the majority of software engineering and knowledge work.”
Carbon sits above it. Google says it’s “built for tasks that test the limits of a model’s intelligence,” such as:
- Complex, multi-step coding projects
- Open-ended scientific research
- Long-horizon agent work that runs over many steps
That’s a sensible split. Most teams don’t need their most expensive model writing routine pull requests. Carbon is the one you bring in when Argon gets stuck.
What the Swarm Feature Changes
The swarm preview is the clearest sign of where Google is aiming. Instead of one agent working through a task step by step, Antigravity can split a single request across several coordinated agents.
If it works well, that matters for long jobs like large refactors or research projects with many parallel threads. Multi-agent setups have mostly been something developers had to build themselves. Google is now putting that orchestration inside its own platform, and Anthropic and OpenAI will feel pressure to do the same.
The Price Is the Real Headline
Cheap frontier pricing is becoming Google’s habit. It did the same thing with Argon, and the gap here is big enough to make finance teams look twice.
There’s a caveat worth taking seriously. A lower per-token price doesn’t guarantee a lower bill. Models can use very different numbers of tokens to finish the same task, so a cheaper model that “thinks” longer can cost more in practice. If you’re evaluating Carbon, compare cost per completed task, not the price list.
How It Stacks Up
The benchmarks Google cited when it first announced Carbon put it among the most powerful models on the market across most metrics. It still trails Anthropic’s Fable 5.2, which launched just a day before Carbon’s wide release.
The usual warning applies. Companies pick the benchmarks that make them look best, so wait for independent testing and your own evals before treating any leaderboard as settled.
Why This Matters
What stands out is how quickly Google got here. For months, frontier AI looked like a two-horse race between Anthropic and OpenAI. With Argon and Carbon launched a month apart, Google has firmly put itself back in that race.
For practitioners, the practical takeaways are:
- Run your evals now. The introductory pricing window is short, so it’s a cheap time to test Carbon on your hardest tasks.
- Route by difficulty. Use Argon for everyday work and save Carbon for the hard problems.
- Watch the swarm preview. If Antigravity’s multi-agent coordination holds up, it could change how teams handle large, long-running tasks.
- Expect price responses. Rivals that cost four to five times more per token will have to justify that premium or cut prices.
The frontier race has three serious players again, and that competition is pushing prices down. Full details on availability and pricing are in the original report from The Verge AI.