Market · AI in practice
Google launches Gemini 4 Neon and points to top spots on tax, legal research and spreadsheet benchmarks
Gemini 4 Neon costs US$ 3 per million input tokens and US$ 15 per million output tokens, with a 1 million-token context window. Google says it leads tax, legal research and finance benchmarks at a 32% lower cost per task than Carbon.
What the sources report
Google placed Gemini 4 Neon between the two models it already had in the family. Gemini 4 Carbon takes the heaviest, longest projects, Gemini 4 Hydrogen handles fast everyday execution, and Neon covers complex workflows that run many times, where speed and cost matter as much as depth. In the API, it is priced at US$ 3 per million input tokens, US$ 0.15 per million cached input tokens, a 95% discount, and US$ 15 per million output tokens.
According to Google, which cites independent evaluations, Neon ranks first on the Tax Agent Bench and the Legal Research Bench, second on the Vals Index at 70.9%, and first on most Surge AI benchmarks across finance, accounting, spreadsheet modeling and complex constraint following.
At its default high thinking level, Google says Neon has a 32% lower cost per task than Carbon and sits 50% to 80% below top competing frontier models on the Vals Index and Agent Arena. It has four thinking levels, so teams can match intelligence and token spend to each workload. The model is available in the paid Gemini API in Google AI Studio, in Gemini for business and, for Google AI Pro and Ultra subscribers, in the Gemini app and Google Antigravity.
My take
In a tax or accounting team, an agent repeats the same routine many times over a month, and the number that drives the purchase is how much each correctly finished task costs. A model that gets close to the frontier on these benchmarks while spending less per task changes that math, and the numbers still need to hold up outside the comparisons Google itself published.