Google DeepMind introduced Gemini 4 Neon this week, the new mid-tier model in the Gemini 4 family that sits between Hydrogen (for fast everyday tasks) and Carbon (for long, complex projects).
According to the official announcement, Neon brings frontier-level intelligence with fast end-to-end task completion across complex software engineering, knowledge work, and science workflows. It scores 61 on the Artificial Analysis Intelligence Index at its xhigh thinking level, ranking third overall and ahead of comparable models.
What Neon offers
The model has a 1 million token context window and is priced at $3 per million input tokens ($0.15 for cached, 95% discount) and $15 per million output tokens. Google says it is roughly three times faster than Carbon and reduces cost per task by around 32% to 60% on third-party benchmarks.
It stands out in agentic coding, ranking first on the Artificial Analysis Coding Agent Index (58.6), with 65.2% on Terminal-Bench 4.1. It also leads on several finance, spreadsheet, legal research, and long-horizon simulation benchmarks such as Vending-Bench 2.
Availability and positioning
Neon is available via the API in Google AI Studio, Gemini for business, for Pro and Ultra subscribers in the Gemini app, and in Google Antigravity. A cybersecurity-focused version is coming soon for the Fairwind Program.
The model replaces Argon in the app selector and offers four thinking levels (Low, Medium, High, and Extra High) to balance speed and depth.
Sources: official Google blog.
By Grok