Google releases Gemini 4 Neon AI model
Google on 18 November introduced Gemini 4 Neon, a new artificial intelligence model aimed at software engineering, knowledge work and science tasks. It is rolling out to developers, businesses and paying subscribers.

Technology company Google on 18 November announced Gemini 4 Neon, a new AI language model in its Gemini 4 family. The company positions it as a fast option for complex software engineering, knowledge work and scientific tasks. Rollout began on the same day.
Pricing and availability
Using the model costs $3 per million input tokens and $15 per million output tokens. Cached input tokens are priced at $0.15 per million, a 95% discount. The context window is one million tokens.
Neon is available in the Gemini API through Google AI Studio, and in Gemini for business. Paying Google AI Pro and Ultra subscribers can use it in Google Antigravity and in the Gemini app. A dedicated cybersecurity version is due to reach vetted defenders within days through the Fairwind Program.
Benchmark results
According to Google, at its highest thinking level Neon scored 61 on the independent Artificial Analysis Intelligence Index, placing third overall, and ranked second on the Vals Index with 70.9%. It took first place on the Artificial Analysis Coding Agent Index with 58.6. On Terminal-Bench 4.1 it scored 65.2%, and on the science-focused Terminal-Bench Science, 70.5%.
The company says that at the default high thinking level, cost per task is 32% lower than for the earlier Carbon model, and roughly 50% to 80% lower than for top competing frontier models.
Safety
Google states that the model keeps the same safeguards against misuse as Carbon. In its internal agent honesty evaluation Neon reached a 99.2% pass rate, and its misaligned action rate on the agentic safety suite was 0.09%. All of these figures are supplied by the company itself.


