Google DeepMind releases Gemini 4 Carbon two weeks after Argon

What happened

Google DeepMind released Gemini 4 Carbon just two weeks after Argon, reaching 59 points on the Artificial Analysis Intelligence Index v5.0 versus Argon's 50. Carbon leads the overall index but trails Opus 5.5 in Claude Code for agentic coding. Selected cybersecurity experts get access starting this week, with broader availability planned after Google collects feedback and adjusts safety guardrails.

Read next

• More changes in today’s briefingGoogle DeepMind announces Genie 4 for real-time world simulation with native audio

Reported by 10 sources

The story so far

Google DeepMind announced Gemini 4 Carbon, which tops Artificial Analysis’s Intelligence Index v5.0 with a score of 59 but is currently available only to selected cybersecurity defense companies, according to The Verge. Google said access would expand to more trusted partners through its Fairwind Program before reaching developers, businesses, and consumers, after further feedback and adjustments to safety guardrails. Artificial Analysis reported that Carbon led four of nine evaluations, while its coding-agent score using Antigravity CLI was 59, behind Claude Opus 5.5 at 62. Google reported cybersecurity scores of 88.2% on SEC-bench Pro, 54.8% on ExploitGym, and 71.0% on CWE-bench, all measured without guardrails. Google also said Carbon was twice as likely as Argon to generate vulnerability-free code and halved unsupported claims compared with Argon. Ars Technica noted that Carbon’s reported benchmark leads over competing frontier models were generally only a few points. Google set introductory pricing at $2 per million input tokens and $10 per million output tokens. The restricted rollout drew criticism from Nathan Lambert, who objected to announcing benchmark scores for a model he could not publicly use. He later softened that criticism, citing difficulties he had heard about in getting a frontier model released.