AI & Agents

Google DeepMind Releases Gemini 4 Neon, a 1M-Context Reasoning Model for Coding Agents

Leave a comment

Google DeepMind has released Gemini 4 Neon, MarkTechPost reports, a reasoning model that sits between Gemini 4 Carbon and Gemini 4 Hydrogen and is aimed at coding agents, knowledge work, and science workflows. Neon is adapted from Carbon. The weights are closed, and it runs on Google’s hosted stack only, with a 1-million-token unified context window, a 64K default output, and up to 1 million tokens through background long decode. Pricing is listed at $3 per million input tokens and $15 per million output tokens. Thinking defaults to high, with an xhigh setting for harder engineering and science work. In Google Antigravity, Neon can orchestrate parallel Hydrogen subagents.

The benchmark numbers in the coverage are Google’s own. MarkTechPost cites a 61 on the Artificial Analysis Intelligence Index, ahead of Claude Opus 5.5 and GPT-6.2 Sol in that table, plus strong long-context and agentic-coding scores. Neon does not win every row. Weights, parameter count, and architecture are not disclosed. The model is rolling out in the Gemini API paid tier, Gemini for business, Google Antigravity for paid Pro and Ultra subscribers, and the Gemini app for those same plans. I have not used Neon.

My take

A mid-tier Gemini 4 model aimed at coding agents is the interesting part, not the spreadsheet of self-reported scores. Closed weights means this is a hosted option, not something you will drop onto a local GPU. If you already live in Google’s agent tools, this is the new middle button to try. I would treat the leaderboard claims as the vendor’s launch post until more independent numbers show up.

God Bless, and May the Source Be with You!

References & Further Reading

Subscribe (RSS) Follow AI & Agents via RSS

Leave a note

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.