Google’s Gemini 4 Carbon Tops AI Math Benchmark, Outperforming OpenAI’s GPT-6.1 Sol

Google’s Gemini 4 Carbon outperforms OpenAI’s GPT-6.1 Sol on the MathArena benchmark but at a significantly higher cost, signaling advanced AI progress relevant to Japanese markets.

Google's Gemini 4 Carbon has claimed the highest position on the MathArena AI math benchmark, surpassing OpenAI's previously leading GPT-6.1 Sol, according to The Decoder. This marks a significant shift in AI performance rankings, with Anthropic’s Fable 5.2 also contributing to displacing GPT-6.1 Sol from the top spot.

The Decoder reported that Gemini 4 Carbon distinguished itself by nearly never giving up on tasks and ignoring time constraints during testing. However, this improved performance comes at a steep price, as the cost of Gemini 4 Carbon is 15 times higher than that of GPT-6.1 Sol.

For Japanese investors and traders closely monitoring AI advancements across FX, crypto, and equities, these developments highlight the increasing sophistication and cost considerations of next-generation AI models that could influence algorithmic trading and market analysis tools.

#newsroom

By Mansa Kane

Tokyo Overnight

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

No spam. Unsubscribe anytime.