What changed
According to Google's November 18, 2026 Gemini API release notes, Gemini 4 Neon targets complex software engineering, knowledge work, and science workflows. The changelog says the model delivers frontier-level intelligence with fast end-to-end task completion. It also supports four thinking levels: `low`, `medium`, default `high`, and `xhigh`.1
The release notes place Gemini 4 Neon alongside three other Gemini 4 models shipped since October 13, 2026. Google's changelog lists Gemini 4 Hydrogen at $1.50 per 1M input tokens and $7.50 per 1M output tokens on the paid tier, plus a free tier. Gemini 4 Carbon carries introductory pricing of $2.00 per 1M input tokens and $10.00 per 1M output tokens through November 30, 2026.1
How it works
Google's Gemini 4 Neon documentation states that the model accepts text, image, video, audio, and PDF inputs while returning text output. In a coding or research session, builders can combine those inputs with code execution, computer use, function calling, and structured outputs. The model card also lists support for agentic video understanding, file search, URL context, and grounding with Google Search and Google Maps.2
For production deployments, the Gemini 4 Neon model page lists three consumption options: the Batch API, Flex inference, and Priority inference. Builders can also use context caching, though the documentation notes that explicit caching is not supported. Google links to a dedicated migration guide and a model comparison page to help teams weigh Neon against Carbon, Hydrogen, and Flash-Lite.2
Google's model card sets out the limits builders should plan for. The default synchronous output limit is 65,536 tokens. Long decode supports up to 1,048,576. The card lists image generation and the Live API as not supported, so voice and image output need another model.2