Google continue the evolution of its range of artificial intelligence with Gemini 4 Neon, its new model intended for complex tasks, software development and data analysis. It promises performance close to the Gemini 4 Carbon, but with up to three times faster execution and a lower cost.

Gemini 4 Neon completes a range that now includes several models with different uses. It sits between the Gemini 4 Hydrogen, designed for quick daily tasks, and the Gemini 4 Carbon, designed for the most demanding thinking jobs. Google thus seeks to offer a compromise between power, speed and cost, especially for users who multiply complex queries.
An AI model designed for development and complex tasks
Gemini 4 Neon stands out for its reasoning abilities and its ability to handle multi-step tasks. Google uses it in particular for programming, debugging entire projects, financial analysis and research work. It can also handle large documents and cross-reference different sources of information.
Its main features are:
- One million token context window: Process large amounts of information in a single conversation.
- Multimodal inputs: text, images, videos and audio.
- Four levels of reasoning: low, medium, high and xhigh to adapt the depth of analysis to the needs.
- Autonomous agents: ability to execute long tasks, use tools, and coordinate multiple agents in Google Antigravity.
Google also announces strong benchmarks in software development, including a score of 65.2% in Terminal-Bench 4.1 and 39% in SWE-Sweep in its Antigravity environment.

Gemini 4 Neon is already third in the Artificial Analysis ranking
The first independent tests are also interesting. Artificial analyzes attributes Gemini 4 Neon scores 61 on his IQ with the Xhigh reasoning level. The model is in third place overall, behind the GPT-6.2 Astra and Claude Fable 5.2. It is noticeably ahead of the Gemini 4 Carbon by one point based on the compared configurations.
The organization also measured the cost and time required to complete its baseline tasks. Gemini 4 Neon averages $6.44 per task, compared to Carbon’s $7.76 at its standard price. It completes these same tasks in about 21 minutes, compared to 36 minutes for its older brother. These results correspond to a reduction of approximately 17% in costs and 40% in time, respectively.
Google’s promise of up to three times the speed and a cost reduced to 60% of Carbon comes, therefore, from initial evaluations communicated by Google. Independent measurements from Artificial Analysis confirm a more moderate improvement, but in the same direction.
Gemini 4 Neon Price and Availability
Gemini 4 Neon is already rolling out to several Google services. Users with a Google AI Pro or Ultra subscription can select it in the Gemini app and Google Antigravity. Businesses can also access it through Gemini for Professionals, while developers have their API in Google AI Studio.
For developers, the price is set at $3 per million input tokens, $15 per million output tokens, and $0.15 per million cached input tokens, a 95% reduction.
For comparison, Carbon is currently $2 in and $10 out thanks to a promotion valid through November 30. Its regular price will then increase to $4 and $20 respectively, making Neon 25% cheaper to use per token.
In the Gemini app, Neon also takes Argon’s place in the model selector, where it becomes the option for in-depth analysis. Hydrogen remains dedicated to day-to-day demands, while Carbon maintains programming tasks and more complex jobs.






