Google releases Gemini 4 Neon model with triple the speed of Carbon

Executive Briefing
“Google has announced its Gemini 4 Neon midtier artificial intelligence model featuring...”
Key Takeaways
- 1.Google announced Gemini 4 Neon as a general-purpose artificial intelligence model built to handle complex coding tasks, multistep agentic workflows, and analytical reasoning.
- 2.According to benchmarks published by the company, Neon completes tasks about three times as quickly as Carbon at roughly 60 percent lower costs, while outperforming the flagship model on certain tasks.
- 3.Independent evaluation results from Artificial Analysis recorded that Neon achieved a score of 61 on the Intelligence Index, ranking third overall.
- 4.Neon is priced at $3 per million input tokens, 15 cents per million cached input tokens, and $15 per million output tokens.
Google has announced its Gemini 4 Neon midtier artificial intelligence model featuring three times the processing speed of the Carbon version at roughly 60 percent lower costs. The general-purpose system is now available to users through the Gemini app and developer platforms.
Key Developments & Primary Facts
Google announced Gemini 4 Neon as a general-purpose artificial intelligence model built to handle complex coding tasks, multistep agentic workflows, and analytical reasoning.
According to benchmarks published by the company, Neon completes tasks about three times as quickly as Carbon at roughly 60 percent lower costs, while outperforming the flagship model on certain tasks.
Independent evaluation results from Artificial Analysis recorded that Neon achieved a score of 61 on the Intelligence Index, ranking third overall.
Neon is priced at $3 per million input tokens, 15 cents per million cached input tokens, and $15 per million output tokens.
Strategic Analysis & Implications
Pricing the model in the middle tier between competing options highlights the effort to secure an advantage as enterprise customers increasingly prioritize cost efficiency.
Outlook & Factors to Watch
The actual performance of the system when deploying large-scale automation workloads outside of standard benchmark datasets still requires further validation.







