Google's new AI model takes the top spot as OpenAI cuts prices

Google's new AI model, Gemini 4 Carbon, now ranks first on the Artificial Analysis Intelligence Index, an independent scorecard of AI models. OpenAI and Anthropic had been ahead for most of the year. OpenAI answered by cutting the price of its mid-tier model in half.

What changed

Google built Gemini 4 Carbon to find and fix security flaws. Moonshot, another AI company, tuned its flagship model for financial research. Other releases this week focused on speed, privacy and size:

• Block and Rime released voice models that process speech directly instead of turning it into text first.
• Block's model runs on a high-end laptop and goes to safety researchers first.
• Rime's lets call centers screen every word before the AI speaks.
• Nvidia and Liquid AI released small, free models that run on PCs and phones without the cloud.

Why it matters

Cheaper AI could lower the cost of tools your business already uses. Small models that run on a phone or PC may keep more of your data off the cloud. Liquid pitches real-time fraud detection as one use for its model.

What is not known yet

The report does not give the new OpenAI price or say how big Google's lead is. It also does not say when Block's voice model will reach the public, or whether software makers will pass savings on to customers.