Gemini 4 Carbon, Google, debuts stronger model, undercutting Claude Opus 5.5 price

The engineering team at Google DeepMind has unveiled Gemini 4 Carbon, a model that boasts superior performance while maintaining the identical initial pricing structure as its predecessor, Gemini 4 Argon.

Koray Kavukcuoglu, who serves as SVP at Google DeepMind and holds the title of chief AI architect at Google, asserted that Gemini 4 Carbon stands as the organization's most powerful model to date, notably excelling in areas such as long-running tasks and cybersecurity applications.

This new model is positioned to directly challenge Anthropic's Claude Opus 5.5, which Google clearly views as its primary competitor. The initial cost is set at $2 for every million input tokens and $10 for every million output tokens, matching the rate of Argon for an unspecified introductory duration. Once this introductory window concludes, the pricing will escalate to $4 per million input tokens and $20 per million output tokens.

Despite some internal naming choices—like "Carbon," which carries an interesting irony given the Chocolate Factory's commitment to a carbon-dioxide-free future, or the chemical symbol "Si" in Gemini 4 Silicon, a subtle nod to "Super Intelligence"—Carbon has demonstrated remarkable technical capabilities.

Its benchmark results are impressive; it has secured the top spots in SEC-bench Pro, ExploitGym, and CWE-bench. Furthermore, it occupies the fourth position on WebDev Arena, trailing behind Opus 5.5, GPT-6.1 Astra, and Sonnet 5.5. In the Arena.ai leaderboard for Best Overall Agent, Carbon currently ranks second, behind Opus 5.5.

Model ratings from biz Artificial Analysis indicate that Google's fresh model tops its newly developed Artificial Analysis Intelligence Index v5, placing second only to Claude Code Opus 5.5 on the Coding Agent Index.

Kavukcuoglu specifically highlighted the security advantages of Carbon, noting that these capabilities will be accessible this week to a select group of cybersecurity specialists, with wider availability planned through the company's Fairwind access program. He stated, "On internal benchmarks, Carbon achieves more than 2x uplift in success rate for generating secure code over Argon." Adding further context, he mentioned, "As another example, the Chrome team has reported that Carbon performed more than 10x better than another lab's large frontier model when generating high quality patches."

Sundar Pichai, the CEO of Google and Alphabet, remarked in a statement accompanying Alphabet's Q3 2026 earnings that "Gemini 4 Argon and Carbon are now leading at the frontier." He also noted, "Overall, our first-party model APIs are now processing nearly 32 billion tokens per minute, more than four times what we processed a year ago."

While Google might aspire to surpass OpenAI and Anthropic in both model performance and usage metrics, the company is already leading in revenue generation. Nate Elliott, a principal analyst at Emarketer, informed The Register via a statement that Google's strong third-quarter earnings were noteworthy, adding, "In a sly nod to the way AI-only companies talk about their revenue, Google bragged about its half-trillion dollar run rate – almost ten times more than either OpenAI or Anthropic."

The rollout for Carbon is scheduled for this week, beginning with certain cybersecurity defenders, though the company intends to broaden the model's accessibility "as soon as possible."

Why it matters

This release directly impacts the competitive landscape for developers choosing AI tooling, especially regarding cost versus capability. The initial pricing strategy suggests Google is aggressively challenging established models like Claude Opus 5.5 in the developer market. Developers should watch how Carbon's performance translates into real-world coding tasks outside of benchmark tests.