Back to all news
AnalysisAlphabet

Google Gemini 4 Carbon — 96% Long-Context · 9.6% SWE-sweep · Leads on Agentic Benchmarks

Published · Updated · By Investus editorial desk (SUNRYU Partners)

Google unveiled Gemini 4 Carbon as the world's most powerful AI model. Long-context score: 96% (rivals ~70%). SWE-sweep: 9.6% vs 4.3%. FrontierMath 2.0: 61.2% vs 55%. Vals Index v5.0: 59 (vs GPT-4 Atlas 67, Claude Opus 5.5 58). Dominates agentic and coding benchmarks; trails GPT-4 Atlas on overall index.

A full English translation of this report is in progress. The analysis below reflects the original Korean research. Read the Korean original →

Google Gemini 4 Carbon — 96% Long-Context · 9.6% SWE-sweep · Leads on Agentic Benchmarks

Sources & further reading

Related articles

Email the morning brief

Disclaimer

This article is editorial research for informational purposes only. It is not investment advice, a recommendation, or a solicitation to buy or sell any security, and does not guarantee any return. Markets carry risk of loss. Verify figures against the primary sources before making decisions.

View Past Sessions