Gemini 4 Carbon Tops Intelligence Index v5

Google’s Gemini 4 Carbon ranks first on the Artificial Analysis Intelligence Index v5, scoring 59, and second to Claude Opus 5.5 on the Coding Agent Index. The model’s Intelligence Index score reflects strong results in complex reasoning, knowledge and scientific tasks. It also improves coding and knowledge work compared with other Gemini models, though it does not lead in every area.

Gemini 4 Carbon is rolling out to selected users and is not publicly available. Google is offering a 50% launch discount with no confirmed end date; the index’s charts and figures use standard pricing for Carbon and Gemini 4 Argon. At high reasoning, Carbon takes the top score in 4 of 9 Intelligence Index evaluations: HLE-Diamond, Terminal-Bench-Science, AA-Omniscience and AutomationBench-AA.

On the Coding Agent Index, Gemini 4 Carbon (high) scores 59 with Antigravity CLI, 3 points behind Claude Code with Claude Opus 5.5 (max) at 62. Carbon ranks second on Terminal-Bench 4.0 at 64% and FrontierCode v1.1 at 51%, behind Claude Opus 5.5 (max) at 56%.

On AA-Briefcase, Carbon scores 1609 Elo, 126 points above Gemini 4 Argon (high) at 1483, but below Grok 4.7 (xhigh) at 1638, Muse Wonder (max) at 1659 and Kimi K3.1 Pro at 1678. Its Presentation score rises 293 Elo to 1598, and Analytical Quality improves 120 points to 1671. At standard pricing, Carbon costs $7.76 per Intelligence Index task, about 23% less than Claude Opus 5.5 (max) at $10.04 and about 1.2x GPT-6 Astra (max) at $6.25. It uses 123k output tokens per task, 29% fewer than Claude Opus 5.5’s 173k. At the same $4/$20 standard pricing, its cost per task is in line with Argon’s $7.68, while Carbon scores 9 points higher on the Intelligence Index.

Carbon scores 54 on AA-Omniscience v1.1, 6 points above Claude Opus 5.5 (max) at 48 and 8 above Argon at 46. Its accuracy rises 10 points over Argon to 58%, while its hallucination rate increases from 4% to 11%; that rate is the second lowest among models, behind Argon, and compares with 50% for Claude Opus 5.5 and GPT-6 Astra (both max). The model has a 1M-token context window, accepts text, image, video and speech input, and produces text. Standard pricing is $4/$20 per 1M input/output tokens; cached input costs $0.20 per 1M at standard pricing, a 95% discount.

Google’s Gemini 4 Carbon ranks first on the Artificial Analysis Intelligence Index v5, scoring 59, and second to Claude Opus 5.5 on the Coding Agent Index. The model’s Intelligence Index score reflects strong results in complex reasoning, knowledge and scientific tasks. It also improves coding and knowledge work compared with other Gemini models, though it does not lead in every area.

Gemini 4 Carbon is rolling out to selected users and is not publicly available. Google is offering a 50% launch discount with no confirmed end date; the index’s charts and figures use standard pricing for Carbon and Gemini 4 Argon. At high reasoning, Carbon takes the top score in 4 of 9 Intelligence Index evaluations: HLE-Diamond, Terminal-Bench-Science, AA-Omniscience and AutomationBench-AA.

On the Coding Agent Index, Gemini 4 Carbon (high) scores 59 with Antigravity CLI, 3 points behind Claude Code with Claude Opus 5.5 (max) at 62. Carbon ranks second on Terminal-Bench 4.0 at 64% and FrontierCode v1.1 at 51%, behind Claude Opus 5.5 (max) at 56%.

On AA-Briefcase, Carbon scores 1609 Elo, 126 points above Gemini 4 Argon (high) at 1483, but below Grok 4.7 (xhigh) at 1638, Muse Wonder (max) at 1659 and Kimi K3.1 Pro at 1678. Its Presentation score rises 293 Elo to 1598, and Analytical Quality improves 120 points to 1671. At standard pricing, Carbon costs $7.76 per Intelligence Index task, about 23% less than Claude Opus 5.5 (max) at $10.04 and about 1.2x GPT-6 Astra (max) at $6.25. It uses 123k output tokens per task, 29% fewer than Claude Opus 5.5’s 173k. At the same $4/$20 standard pricing, its cost per task is in line with Argon’s $7.68, while Carbon scores 9 points higher on the Intelligence Index.

Carbon scores 54 on AA-Omniscience v1.1, 6 points above Claude Opus 5.5 (max) at 48 and 8 above Argon at 46. Its accuracy rises 10 points over Argon to 58%, while its hallucination rate increases from 4% to 11%; that rate is the second lowest among models, behind Argon, and compares with 50% for Claude Opus 5.5 and GPT-6 Astra (both max). The model has a 1M-token context window, accepts text, image, video and speech input, and produces text. Standard pricing is $4/$20 per 1M input/output tokens; cached input costs $0.20 per 1M at standard pricing, a 95% discount.