Gemini 4 Carbon is the new #1 on the Artificial Analysis Intelligence Index v5
Points and comments are a snapshot, not live.
Gemini 4 Carbon tops Artificial Analysis Intelligence Index v5, trails Claude Opus 5.5 on coding agents.
Google's Gemini 4 Carbon (high reasoning) scores 59 on the Intelligence Index v5, leading in 4 of 9 evaluations including HLE-Diamond and AA-Omniscience. It ranks second on the Coding Agent Index at 59, behind Claude Opus 5.5 (max) at 62. On AA-Briefcase, Carbon scores 1609 Elo, a 126-point improvement over Argon but behind models like Kimi K3.1 Pro (1678). Carbon costs $7.76 per Intelligence Index task, ~23% less than Claude Opus 5.5 but ~1.2x GPT-6 Astra. It has a 1M token context window, multimodal input, and standard pricing of $4/$20 per 1M tokens with a 50% launch discount.
The model is not yet publicly available, rolling out to selected users. Google offers a 50% launch discount with no confirmed end date. Carbon uses 123k output tokens per task, 29% fewer than Claude Opus 5.5.
What commenters are saying
Commenters are sharply divided. Many dismiss Google's announcement as a PR stunt for investors, noting the model is not publicly available and that Argon had limited impact. One top commenter says the release coincided with Google's earnings call and missed targets, with $GOOG down 13%. Another shares personal experience with Argon, citing dreadfully low limits (5 messages on medium triggers a 4-hour cooldown) and active hallucination issues. A smaller group argues Gemini is strong for knowledge work and enterprise, with one commenter describing a large mortgage firm entirely using Gemini to find unaccounted cases.
Several point out Google's pattern of announcing models long before release, contrasting with other labs that announce closer to launch. Some attribute this to DeepMind's cultural DNA of not releasing and to internal performance review deadlines.