Gemini 4 Carbon vs GDPx: Second Place and Significantly Lower Price

The new GDPx benchmark from SemiAnalysis ranked Google's Gemini 4 Carbon in second place, just behind Anthropic's Claude Opus 5.5. According to 9to5google.com, however, Google's new model significantly outperforms the competition in terms of performance-to-price ratio.

The benchmark includes 206 tasks across various fields such as finance, materials science, cybersecurity, software development, law, biology, and more. Claude Opus 5.5 achieved the highest score at a cost of around $51 per task, while Gemini 4 Carbon was priced at just over $16 per task.

In third place was OpenAI's GPT-6 Astra, which scored lower but was even more expensive than the previous two models, costing around $54 per task.

GDPx Benchmark Results Excluding Certain Domains

If cybersecurity, biology, and alignment are excluded from the results, the ranking changes. GPT-6 Astra takes the top spot, followed by GPT 6.1 Sol and Claude Sonnet 5.5, which surpass Gemini 4 Carbon. SemiAnalysis noted that this better reflects their "subjective impressions" of the individual models.

The existing model, Gemini 4 Argon, finished eighth in both rankings, surpassing Meta's Muse Wonder but falling behind Wonder Full. In cybersecurity, Google's models outperform those from OpenAI and Anthropics, but according to the benchmark, they lag behind Meta's Muse models, which were the best "by a meaningful margin."

Announcement of Gemini 4 Carbon and Release Date

Google announced Gemini 4 Carbon on October 28, boasting of "significant gains," but has yet to confirm a public release date. For most customers, the model is labeled as "coming soon." Google's own benchmark shows that Gemini 4 Carbon outperforms GPT-6 Astra and Opus 5.5 in many aspects, which differs from the results of SemiAnalysis.