On November 18, U.S. time, Google announced "Gemini 4 Neon," a new high-performance AI model capable of handling complex tasks. It excels at programming, analyzing large volumes of data, and scientific research. Gemini 4 Neon will gradually become available through Gemini apps like Google AI Pro and Ultra, as well as via APIs.

Neon's key feature is its balance of performance, processing speed, and cost. Google's Gemini 4 series includes "Carbon," designed for deliberate processing of difficult tasks, and "Hydrogen," which quickly handles everyday operations. Neon is positioned between these two, aiming to process tasks requiring advanced reasoning in a relatively short time and at a low cost.

In the "Artificial Analysis Intelligence Index," which evaluates overall AI capabilities, Neon scored 61 points, securing the third overall position. According to Google's comparison chart, Anthropic's "Claude Opus 5.5" scored 59 points, and OpenAI's "GPT-6.2 Sol" scored 57 points, with Neon outperforming both models.

On the other hand, in "Terminal-bench 4.1," a test where AI performs actual programming tasks, Claude Opus 5.5 slightly surpassed Neon with 65.8% compared to Neon's 65.2%.

API pricing is $3 per million input tokens and $15 per million output tokens. According to Google, with standard inference settings, the cost per task can be reduced by 32% compared to Carbon.

Meet with AI