The latest comparison of frontier AI models highlights a growing focus on both performance and operating costs. Google’s Gemini 4 Carbon has secured the 2nd position in SemiAnalysis’ new GDPx benchmark, which evaluates how effectively AI models handle agentic tasks. Although Anthropic’s Claude Opus 5.5 achieved a higher score, Gemini 4 Carbon emerged as the most cost-efficient model among the top 5.
Claude Opus 5.5 leads SemiAnalysis’ GDPx benchmark
In a post on X, SemiAnalysis shared results from its new benchmark, which assesses the agentic capabilities of leading AI models from OpenAI, Google, Anthropic, Meta, xAI and other AI companies.
Anthropic’s Claude Opus 5.5 topped the rankings with a score of 12 percentage points. Google’s Gemini 4 Carbon followed closely with 10.9 percentage points.
Despite securing the 2nd position, Gemini 4 Carbon recorded the lowest cost per task among the top 5 models. It cost $16.29 per agentic task and generated 594,000 output tokens. In comparison, Claude Opus 5.5 cost an average of $51.41 per task while generating 548,000 output tokens.
OpenAI’s GPT-6 Astra ranked 3rd with 9.6 percentage points, costing $53.99 per task for 275,000 output tokens. Meta’s Muse Wonder Full followed with 9.0 percentage points.
Anthropic’s Claude Fable 5.1 scored 8.5 percentage points and recorded the highest cost per task at $67.16, with 418,000 output tokens. Google’s Gemini 4 Argon also featured in the rankings with 7.3 percentage points. However, it was more cost-efficient than Gemini 4 Carbon, with a cost of $12.79 per task.
Gemini 4 Carbon’s cybersecurity capabilities
Google introduced Gemini 4 Carbon in October with enhanced cybersecurity capabilities. However, access is currently restricted to a “subset of cyber defenders”. The company plans to gradually extend availability to “additional trusted partners” through its Fairwind Program.
Google also claims that internal testing showed Gemini 4 Carbon achieved more than a 2x improvement in the success rate of generating secure code compared with Gemini 4 Argon.
The model was launched late last month, just 2 weeks after Google introduced Gemini 4 Argon, another AI model designed for agentic capabilities.
Also read: Viksit Workforce for a Viksit Bharat
Do Follow: The Mainstream LinkedIn | The Mainstream Facebook | The Mainstream Youtube | The Mainstream Twitter
About us:
The Mainstream is a premier platform delivering the latest updates and informed perspectives across the technology business and cyber landscape. Built on research-driven, thought leadership and original intellectual property, The Mainstream also curates summits & conferences that convene decision makers to explore how technology reshapes industries and leadership. With a growing presence in India and globally across the Middle East, Africa, ASEAN, the USA, the UK and Australia, The Mainstream carries a vision to bring the latest happenings and insights to 8.2 billion people and to place technology at the centre of conversation for leaders navigating the future.