Google Gemini 4 Carbon is here: Is it better than OpenAI GPT-6 and Anthropic Claude?
Google has introduced Gemini 4 Carbon, its latest frontier artificial intelligence model, on October 28, 2026. The new model focuses on coding, cybersecurity, complex reasoning and tasks that can take hours or even days to complete. Google claims Carbon is its most capable AI model yet, arriving just two weeks after the launch of Gemini 4 Argon. The release adds fresh competition for OpenAI and Anthropic, which have been developing increasingly capable AI systems.
The headline figure comes from Artificial Analysis. Gemini 4 Carbon has scored 59 points on the Intelligence Index v5.0, ahead of OpenAI’s GPT-6 Astra at 57 and Anthropic’s Claude Opus 5.5 at 58. Google’s published results show improvements in coding, cybersecurity testing and handling large amounts of information. Access remains restricted to selected cybersecurity organisations during the initial rollout.
:
Gemini 4 Carbon vs OpenAI and Anthropic: Benchmark comparison
According to Google’s published benchmark results, Carbon leads several important AI tests.
| Benchmark | Gemini 4 Carbon | GPT-6 Astra | Claude Opus 5.5 |
|---|---|---|---|
| AA Intelligence Index v5.0 | 59 | 57 | 58 |
| Terminal-Bench 4.0 | 65.2% | 59.1% | 59.6% |
| FrontierSWE 2 | 61.2% | 65.5% | 62.3% |
| Agents’ Last Exam v1.1 | 77.4% | 73.1% | 71.6% |
| GraphWalks long-context test | 96.2% | 71.8% | 66.8% |
| CWE-bench cybersecurity | 71.0% | 68.0% | 67.0% |
:
What makes Gemini 4 Carbon different?
Cybersecurity is one of Carbon’s main focus areas.
Google reports scores of 88.2% on SEC-bench Pro, which tests vulnerability discovery, and 54.8% on ExploitGym, which evaluates exploitation tasks.
The company says Carbon can work on longer software engineering problems and produce more secure code. Google reports fewer unsupported claims compared with Argon, based on internal testing. These claims still need wider independent evaluation.
:
Gemini 4 Carbon pricing and availability
Google has announced introductory API pricing of $2 per million input tokens and $10 per million output tokens.
Using exchange rate of ₹95 per dollar, that works out to approximately ₹190 for input and ₹950 for output per million tokens, before taxes.
Cached input tokens receive a 95% discount.
Carbon is initially available to selected cybersecurity experts, with broader access planned through Google’s Fairwind programme. The company has not announced a firm date for general availability.