FULL STORY

Gemini 4 Neon: From Leak to Launch in Days

Gemini 4 Neon surfaced across Gemini code, Cloud billing, and GCP quota pages before being officially listed in Vertex AI within days, positioned as a fast model approaching Carbon's quality.

2026-11-18 ~ 2026-11-19 · 4 episodes · 20 posts

Episode 1 · Unreleased Gemini 4 Neon Appears in Google Cloud Billing, Priced $3/$15 (2026-11-18, 7 posts)

The unreleased model codenamed Gemini 4 Neon is showing up at an accelerating pace: after being spotted in Gemini web frontend code on November 16 (as relayed by @ExistingHat1064), on November 18 multiple users and model-tracking accounts reported that the model had appeared in Google Cloud's SKU/billing catalog, with pricing revealed — widely seen as a sign of an imminent launch.

Confirmed

  • The model now appears in the Google Cloud SKU/billing catalog, with prices officially in effect (as relayed by @koltregaskes from the model-tracking account aitrackerbot, effective from 3 AM ET)
  • Pricing is $3 per million input tokens and $15 per million output tokens
  • Per @testingcatalog, the accompanying image shows the Gemini API price sheet also includes tiered audio/video pricing
  • @testingcatalog notes this process mirrors the lead-up to previous Gemini model launches

Not Yet Confirmed

  • A Reddit tip relayed by @likeastar20 suggests the model could launch as early as today, though it is explicitly flagged as unverified
  • Google has made no official announcement

Why It Matters

The appearance of a billing catalog is typically one of the final preparation steps before Google launches a new model. If the pricing holds, the $3/$15 rate will directly shape how developers weigh Gemini 4 Neon against competitors on cost, and the launch window may be very close.

Episode 2 · Gemini 4 Neon surfaces in GCP quota page as Google announcement nears (2026-11-19, 6 posts)

On Nov 19, multiple leaks pointed to an imminent Google release of Gemini Neon. The model codename 'gemini-4-neon' appeared in Google Cloud's quota page after surfacing in AI Studio, though claims of Opus 5.5-level capability remain unconfirmed.

Episode 3 · Gemini 4 Neon Spotted in Google AI Studio Alongside Mystery 'Cobalt' Codename (2026-11-19, 3 posts)

Google AI Studio's source code revealed a "gemini-4-neon" model ID along with a mysterious new "Cobalt" codename, with users confirming the deployment and speculating an official launch is imminent.

Episode 4 · Google Launches Fast-Inference Gemini 4 Neon (2026-11-19, 4 posts)

Google has launched Gemini 4 Neon, a fast-inference model built on Gemini 4 Hydrogen, now GA on Vertex AI and the Gemini API paid tier at $3/$15 per million tokens. It offers near-Carbon quality with Flash-level speed and cost.