Gemini 4 Neon Is Here — and Argon Is Already Gone

Hand typing on a backlit keyboard in a dark room lit by blue and red screen glow

Google has launched a new Gemini model — and retired an old one on the same day. According to Google, Gemini 4 Neon arrived on Tuesday as the fourth model in the Gemini 4 family in about seven weeks, built for complex coding, analytical reasoning, and enterprise work. The surprise is what disappeared alongside it: Argon, the mid-tier model Google introduced only weeks ago, is gone from the Gemini app’s model picker — along with the older Gemini 3.1 Pro.

Neon sits between Hydrogen, the fast everyday model, and Carbon, the flagship built for the hardest long-horizon projects. It carries a 1 million-token context window and four thinking levels that let users dial intelligence and token spend up or down per task, and at its default High setting Google says it costs 32% less per task than Carbon. It is rolling out now in the paid Gemini API, Gemini for business, Google Antigravity, and the Gemini app for Pro and Ultra subscribers.

The model picker just got a rewrite

The most telling change is not in the model — it is in the menu. As 9to5Google reports, the Gemini app no longer asks subscribers to pick a model by name. It asks how hard the job is: Auto adapts to your needs, Hydrogen handles everyday help, Gemini 4 Neon takes deeper analysis, and Carbon is reserved for code and complex work. Argon is gone outright, and 3.5 Flash-Lite has been folded into Auto.

That is a quiet admission about where this race is heading. Model names now have the shelf life of a phone update. What Google is really selling is a set of effort tiers with a price attached to each — and Neon is the new middle rung, the one most paying users will live on.

Faster than Carbon, cheaper than the rivals

On the numbers, Gemini 4 Neon is a strong middleweight. It scores 61 on the Artificial Analysis Intelligence Index at its highest thinking level, third overall behind larger, slower models, and tops the Artificial Analysis Coding Agent Index at 58.6. It posts 65.2% on Terminal-Bench 4.1 and 46.2% on FrontierCode 2.0, and leads on the Vals Index’s legal and tax benches plus Andon Labs’ Vending-Bench 2 — a long-horizon simulation of running a business — with a $16,843 score.

Notably, Google’s own benchmark table puts Gemini 4 Neon head-to-head with named rivals — Claude Opus 5.5 and GPT-6.2 Sol — rather than against a vague “frontier” standard. As Digital Trends notes, Neon leads on most of the evaluations Google lists, including the overall intelligence index, where it scores 61 against 59 for Opus 5.5 and 57 for GPT-6.2 Sol.

The pricing is where the strategy shows. Gemini 4 Neon costs $3 per million input tokens and $15 per million output tokens, with cached input at $0.15. Carbon and Argon cost $2 and $10 under a promotional rate that ends November 30, when they jump to $4 and $20 — leaving Neon 25% cheaper than Carbon’s standard price from December 1. Google is not just launching a model; it is setting the price of its permanent middle tier.

Why this matters

Watch the cadence, not just the model. Four Gemini 4 models in seven weeks means Google can now ship, measure, and kill a model faster than most developers can build on it — and Argon’s overnight disappearance proves it. If you built workflows around Argon, you are being moved to Gemini 4 Neon whether you asked or not.

That churn is the real story. The frontier-model race has become a subscription business: what matters is no longer which model you use, but which effort tier you pay for — and Google just redrew those tiers while most users were not looking. Neon is the tier Google expects you to live in. The question is whether the next model retires it just as quickly.

Meanwhile in AI Tech: Google’s Argon launch promised flagship coding at half price — now the model itself is gone from the app. Meanwhile in AI Tech: OpenAI’s GPT-6.2 Sol is the rival Neon is benchmarked against, and the price war between the two is only getting sharper.

Written by
Ryan covers artificial intelligence and enterprise tech — from foundation models and AI chips to the business of machine intelligence. He tracks model releases, funding rounds, and the policy moves shaping the AI industry.