Gemini 4 Carbon Dethrones Claude Opus 5.5 With a Bang
Trending Topics Jakob Steinschaden ● Covered by 4 sources
Google’s Gemini 4 Carbon just took the top AI ranking spot from Claude Opus 5.5. Bad timing for Anthropic, which is heading toward an IPO.
Based on reporting by Trending Topics, Jakob Steinschaden — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Google has pushed Gemini 4 Carbon to the front of the pack, at least for now. Artificial Analysis puts it at 59 points in its Intelligence Index v5, one point ahead of Anthropic’s Claude Opus 5.5. The timing lands awkwardly for Anthropic, whose IPO paperwork is already with US regulators and whose Nasdaq debut is expected in the coming weeks.
Carbon’s strength is broad, but not absolute. It tops four of the index’s nine individual tests, including complex reasoning, knowledge and scientific work. Against Gemini 4 Argon, the model that started rolling out to paying users two weeks ago, Carbon adds 9 points. The biggest jump is in coding, where Google has spent a long time playing catch-up.
That catch-up is still partial. In Artificial Analysis’ Coding Agent Index, Carbon scores 59 points with the Antigravity CLI, just behind Claude Code with Opus 5.5 at 62. It also comes second on Terminal-Bench 4.0 and FrontierCode v1.1. On reliability, Carbon looks stronger: its AA-Omniscience score rises to 54 from 48 for Opus 5.5 and 46 for Argon, while accuracy improves by 10 points over Argon to 58 percent. The hallucination rate rises from 4 to 11 percent, though Artificial Analysis still says that is the second-lowest among all models.
Google is also pitching Carbon as a cyber defense tool, not just a general-purpose model. The company says it leads on finding vulnerabilities, exploiting them and fixing them, and claims the Chrome team got patches from it that were more than ten times better than from another lab’s frontier model. Carbon also appears cheaper to run per task than Opus 5.5 at standard prices, though it still costs more than GPT-6 Astra on that measure.
But the crown comes with strings attached. Carbon is only available to selected users right now, starting with cyber defense teams and other partners in Google’s Fairwind Program. Google is also still running it through US government pre-release testing. So yes, Gemini 4 Carbon has the headline. It just doesn’t yet have the whole market.
My take — AI-written commentary, not fact-checked reporting
This is exactly how AI rankings should be read: useful, noisy, and very good at ruining launch plans. Google gets the bragging rights, but a one-point lead on a freshly rebuilt benchmark is not a coronation, it’s a marketing memo with a stopwatch. The real pattern is simpler: the frontier is now a revolving door, and nobody gets to sit on the throne for long.
Read more about this at: Trending Topics