Summary
Google has announced Gemini 4 Carbon, a new frontier model the company positions as its strongest yet, with state-of-the-art cybersecurity and long-horizon capabilities. This video from Matthew Berman walks through Google’s announcement blog and the published benchmark results to ask whether the claims hold up.
Carbon takes the top spot on the Artificial Analysis Intelligence Index V5 with a score of 59, ahead of Claude Opus 5.5 (58) and GPT-6 Astra (57). The video covers results on Terminal Bench, OS World, Agents Last Exam, SEC Bench Pro, and Vending Bench, comparing Carbon against Gemini 4 Argon, which launched only two weeks earlier, as well as Claude Fable 5.1. Opus 5.5 still leads on computer-use tasks, with Carbon in second place at 81.5.
The rollout is deliberately staged. Carbon goes first to a subset of cyber defenders through Google’s Fairwind program, while Google takes part in the US government’s voluntary pre-release testing process, before broader availability to developers, businesses, and consumers. Introductory pricing is $2 per million input tokens and $10 per million output tokens, rising to $4 and $20 afterward, which undercuts many models at this performance tier. Berman also discusses Google’s slow gap between announcement and release, and what the rapid Argon-to-Carbon pace says about Google’s momentum against OpenAI and Anthropic.
📺 Source: Matthew Berman · Published October 29, 2026
🏷️ Format: News Analysis
