Google Deepmind’s Gemini 4 Argon AI Surpasses 95% on ARC-AGI-3 Benchmark

Google Deepmind's Gemini 4 Argon AI model scored 97.5% on the ARC-AGI-3 benchmark, becoming the second lab after OpenAI to surpass 95%, though it shows weaknesses in math and software engineering.

Google Deepmind's latest AI model, Gemini 4 Argon, scored an impressive 97.5% on the ARC-AGI-3 benchmark, making Google the second company after OpenAI to exceed the 95% threshold, according to The Decoder. This achievement highlights Google's growing presence in advanced artificial general intelligence research.

However, The Decoder also reports that Gemini 4 Argon's performance drops significantly to 49.8% when evaluated using the standard test harness. Additionally, the model currently lags behind its competitors in specific areas such as math and software engineering.

For Japanese investors and tech markets, this development underscores the intensifying global competition in AI innovation, which could influence technology stocks and the adoption of AI-driven solutions across industries.

#newsroom

By Mansa Kane

Tokyo Overnight

Japan markets, decoded. Get the brief in your inbox.

No spam. Unsubscribe anytime.