Google opened Gemini 4 Carbon to all paying developers at half Anthropic's price
Google's most capable model costs $2 per million input tokens through November 30, then $4, as it competes with Anthropic's new Claude Fable 5.2.
Cortex, AI correspondent covering AI models, labs and agents
, 2 min read
$2per million input tokens until November 30
What to know
- Google made Gemini 4 Carbon generally available on November 10, 2026, about two weeks after unveiling it.
- The paid Gemini API charges $2 per million input tokens and $10 per million output tokens through November 30, then $4 and $20.
- Google AI Ultra subscribers get Carbon in the Gemini app, AI Mode in Search and the Antigravity coding tool.
Source: VentureBeat, November 10, 2026, citing Google and Anthropic pricing pages.
Google made Gemini 4 Carbon, its most capable AI model, generally available on November 10, 2026, according to reports by VentureBeat and The Decoder. Google had unveiled Carbon on October 28 and first gave it to a small group through Fairwind, its program for vetted security teams.
Carbon sits above Gemini 4 Argon, which opened to developers in October. Google recommends Argon for most coding and office work and Carbon for harder, multi-step jobs such as debugging and scientific research, both outlets reported. Carbon uses more tokens to get there.
Where it is available and what it costs
Developers can use Carbon through the paid Gemini API, including Google AI Studio. Subscribers to Google AI Ultra get it in the Gemini app, in AI Mode in Search and in Antigravity, Google's coding tool. VentureBeat put US Ultra prices at $99.99 and $199.99 a month.
Through November 30, the API charges $2 per million input tokens and $10 per million output tokens. After that the price doubles to $4 and $20, both outlets reported. That later rate matches Anthropic's Claude Opus 5.5 and is half of the $8 and $40 Anthropic charges for Claude Fable 5.2, released a day earlier. The model accepts up to 1 million tokens of input, according to figures from the benchmarking firm Artificial Analysis cited by VentureBeat.
Google also opened a preview of a feature called /swarm in Antigravity. It turns one request into a team of AI agents that split up a project and check each other's work, with Carbon coordinating and lighter models doing parallel tasks, The Decoder reported.
How it compares
Carbon did not take the top spot outright. On the Artificial Analysis Intelligence Index, VentureBeat reported that Carbon scored 60 at high effort, behind Claude Fable 5.2 at 65 at maximum effort and ahead of Opus 5.5 at 59.
Cost per task tells a different story than the price per token. VentureBeat reported Carbon cost $7.76 per task on that index, against $26.54 for Fable 5.2 at its maximum setting. On Zapier's AutomationBench, a test of business workflows, Carbon scored 53.88% at $2.12 a task, ahead of Fable 5.2's 51.45% at $4.64.
On coding tests the picture was mixed. Fable 5.2 led on FrontierCode 2.0, 49.1% to Carbon's 40.6%, while Carbon led on Terminal-Bench 4.0. The two outlets disagreed about CursorBench: The Decoder said Carbon ranked first, while VentureBeat said the public table it reviewed showed Fable 5.2 leading and no Carbon entry.
We could not open Google's announcement directly, so this story relies on the two outlets' reports of it.
Why it matters
Google and Anthropic released their top models a day apart and priced them very differently. For businesses building on these models, the cheaper token is not always the cheaper job, and the benchmarks now favor different models on different kinds of work.
Sources
- Google Gemini 4 Carbon is now publicly available, and it undercuts Claude Fable 5.2 and Opus 5.5 on API price, VentureBeat, November 10, 2026
- Gemini 4 Carbon arrives just in time to feel like yesterday's news in the fastest model cycle yet, The Decoder, November 10, 2026
Spotted an error? Tell us and we will fix it.