Gemini 4 Neon went from rumor to announcement in two days: on November 16, TestingCatalog found the name gemini-4-neon in Gemini web code; in the early hours of the 18th, Google Cloud's billing catalog leaked the $3/$15 price; that evening, Google announced it officially and shipped it the same day. It is the fourth Gemini 4 model in four weeks, and the family's first mid-tier — above the fast Hydrogen, below the flagship Carbon, pitched as "Carbon-level quality at Flash speeds". This article reconstructs the whole sequence and attributes every claim.
On November 16, 2026 at 20:32 UTC, TestingCatalog — which tracks unreleased features — posted that it had found a reference to gemini-4-neon in Gemini web code; the model was not yet available. The author added a guess and explicitly labelled it pure speculation: Neon might be a new Flash Lite model, the lightest and cheapest tier in the family. The post gave no price, date or API identifier.
The guess turned out to be wrong — Neon is a mid-tier model at twice Hydrogen's price — but the finding itself was real: multiple checks confirmed the string existed. A post on the Chinese forum Linux.do then mentioned it as gemini-4-neon, and the tracker account NPowerUser repeated it on November 17 without adding any source of its own. Notably, the name "Neon" had already appeared once before, on May 5, 2026: 9to5Google's APK Insight team found traces of a subscription plan codenamed Neon in a macOS Gemini app update, tentatively called Google AI Ultra Lite (between the $20/month Pro and the $250/month Ultra). That was a subscription codename, not a model; Google later reused the name for this model.
In the early hours of November 18, Google Cloud's billing catalog (SKU catalog) listed Neon's price ahead of the announcement: $3 in / $15 out per 1M tokens, $0.15 cached input, effective 3 AM ET. The AI Tracker account found it first, corroborated by TestingCatalog and others. The same day, Antigravity CLI 1.5.11 added a gemini-pro-orchestrator-neon orchestrator variant. AI Frontier Post filed that morning with the headline making clear this was a rumor, not a launch: no model card, no API changelog entry, no announcement from Google — "everything above is paperwork and plumbing, not a product".
By that evening the evidence had flipped completely: the documentation page that returned a 404 in the morning now described gemini-4-neon in full, the DeepMind model card was live, and Neon appeared in the Vertex AI Model Garden and Google AI Studio, callable through the API. AI Frontier Post's recap was candid: in the morning it was a rumor; by evening it was an official launch — because that was, by then, the truth.
On November 18, 2026, Google officially announced Gemini 4 Neon on blog.google (by Koray Kavukcuoglu, SVP Google DeepMind / Google Chief AI Architect). The headline numbers: 61 on the Artificial Analysis Intelligence Index (xhigh effort) for third place; 58.6 on the AA Coding Agent Index at No. 1; 70.9% on the Vals Index at No. 2; 1336 on the Agent Arena Index at No. 2; $3 in / $15 out per 1M tokens, $0.15 cached input; a 1M context; four thinking levels (low / medium / high / xhigh). The positioning in one sentence: Carbon takes the hardest long-horizon projects, Hydrogen handles fast everyday execution, and Neon balances depth and efficiency for high-frequency complex software engineering, knowledge work and science.
Access shipped with the announcement: the Gemini API paid tier (AI Studio, Android Studio), Gemini for business (API and model dropdown), Google Antigravity (paid Pro / Ultra subscribers), and the Gemini app (Pro / Ultra subscribers). The DeepMind model card went live the same day, confirming Neon is "adapted from Carbon", with a March 2026 knowledge cutoff and an 18-test comparison table (all Google's own runs — see our benchmarks article). Gemini 4 Neon for cyber, the cybersecurity variant, waits a few more days for Fairwind Program trusted defenders.
One of the most-discussed details of launch night did not come from Google but from 9to5Google: the Gemini app's model picker now has four entries — Auto, Everyday help (4 Hydrogen), Deeper analysis (4 Neon), Code & complex work (4 Carbon) — with Neon taking Argon's place and 3.1 Pro going away. Google's official blog mentions none of the picker change, Argon's fate or 3.1 Pro's retirement, and rrtechpulse noted the same gap in its own check. This site reports the picker change as 9to5Google's reporting, not as Google's official wording, and will update it when Google confirms it or the picker actually rolls out.
Neon's rumor status is closed, but several things remain to be followed up: the full Neon Safety & Security Report is not yet published (the model card says "in the coming days"); the Vertex AI listing so far rests on tracker reports, while Google's blog only says "Gemini for business" without naming Vertex AI separately; LiveBench 2.0 has no Neon score; one day after launch there is no independent hands-on test anywhere; and mainland Chinese first-tier tech media have not published depth coverage yet. This site's 8-hour tracking loop will watch each of these.
Reference sources: this article is an original rewrite produced by this site; facts and figures were checked against Google's official blog (Nov 18, 2026), 9to5Google (Nov 18, 2026) and AI Frontier Post's rumor-to-launch timeline (Nov 18, 2026)
← Back to the Gemini 4 Neon topic page