Home  /  Blog  /  Gemini 4 Neon released
Model release

Gemini 4 Neon released: $3/$15, no introductory price — and no League Score yet.

Gemini 4 Neon was released on November 18, 2026 and enters our database today at $3/$15 per million tokens, cached $0.15 — a new tier adapted from Carbon, positioned between Hydrogen and Carbon. It is the family’s most expensive model per token until November 30 and its second-cheapest from December 1, without its own price changing. This post prices the family, works the cached maths, and states why it carries no League Score yet.

By AI Pro League · Published

Gemini 4 Neon released — $3/$15 per million tokens, flat with no introductory price, and unranked in the AI Pro League database

Gemini 4 Neon is the newest entry in our model database — one of 67 models we track, 54 of them scored and 13 unranked. Our November 19, 2026 changelog records the release from Google’s official announcement (blog.google, read today) and the Gemini API pricing page (ai.google.dev, read today): released Nov 18, 2026 at $3.00 input / $15.00 output per million tokens, cached input $0.15, via Google AI Studio, Gemini API, for complex coding, analytical reasoning and enterprise workflows. What follows is computed from the database at build time.

Gemini 4 Neon: the release record, dated and priced

Gemini 4 Neon in the AI Pro League database (prices checked 2026-11-19)
MeasureGemini 4 Neon
Release dateNov 18, 2026
Input / output per 1M tokens$3.00 / $15.00
Cached input per 1M tokens$0.15 (95% below fresh input)
Blended rate (3:1 input:output)$6.00
Context window / max output1,048,576 / 1,048,576 tokens
InputsText, Image, Audio, Video, Pdf
Knowledge cutoffMarch 2026
League ScoreUnranked — no tracked benchmark yet

Two record details deserve plain words. The maximum output in the database — the full 1,048,576-token window — is the background long-decode figure from Google’s model documentation, recorded in our changelog; a single synchronous segment is capped at 65,536 output tokens, so the headline window and the single-call output are different numbers. And unlike its siblings, Neon carries no introductory price: its card is flat, which is what makes the family table below move around it on December 1. The full rate card sits in our AI model pricing index.

The Gemini 4 family, priced from one database

The Gemini 4 family in the AI Pro League database — standard workload is 10M input + 2M output tokens; cached-heavy run is 8M cached + 2M fresh input + 2M output; sibling rates are their current introductory card
Model$/1M in / out / cachedBlendedWorkload costCached-heavy costLeague Score
Gemini 4 Hydrogen$1.50 / $7.50 / $0.075$3.00$30.00$18.6085.0 (#14, 40% coverage)
Gemini 4 Neon (new)$3.00 / $15.00 / $0.15$6.00$60.00$37.20Unranked (0% coverage)
Gemini 4 Carbon$2.00 / $10.00 / $0.10$4.00$40.00$24.8097.3 (#4, 40% coverage)
Gemini 4 Argon$2.00 / $10.00 / $0.10$4.00$40.00$24.8086.3 (#12, 40% coverage)

Read the ordering carefully, because it inverts. Today Neon is the expensive tier: its standard workload costs $60.00 — 10 × $3.00 + 2 × $15.00 — exactly twice Gemini 4 Hydrogen’s $30.00 and 50% above the $40.00 that Gemini 4 Carbon and Gemini 4 Argon charge at introductory rates, and its blended $6.00 is the family’s highest rate. The family it joins is a single-component family: Carbon’s 97.3 (#4), Argon’s 86.3 (#12) and Hydrogen’s 85.0 (#14) each rest on one recorded LiveBench component at 40% coverage — the standings live in the AI model rankings, and the siblings can be set against each other on the model comparison pages.

December 1: the step that reprices Neon without touching it

On December 1, 2026, Carbon and Argon step from their introductory $2.00 / $10.00 to their standard $4.00 / $20.00 per million tokens, cached $0.10 to $0.20, per the pricing page read today. Their standard workload doubles, from $40.00 to $80.00. Neon’s rate does not move — and that stillness is the reprice: at $60.00, its workload goes from 50% above its siblings to 25% below them overnight, and its headline rates sit exactly 25% below the standard card on every line — $3.00 against $4.00, $15.00 against $20.00, $0.15 against $0.20. Any family budget comparison written this month should say which side of December 1 it prices; the step itself is worked for Carbon in our Gemini 4 Carbon first-score analysis.

The cache tells the same story on the run agent loops actually execute — 8M cached input + 2M fresh input + 2M output. Neon costs $37.20 — 8 × $0.15 + 2 × $3.00 + 2 × $15.00 — because cached tokens cost 95% less than fresh input on its card. Hydrogen runs the same mix for $18.60; Carbon and Argon run it for $24.80 today and $49.60 from December 1. Neon’s cached rate is exactly double Hydrogen’s and exactly 25% below the siblings’ December cached rate — the field-wide pattern behind those gaps is priced in our prompt caching pricing analysis.

Why Gemini 4 Neon has no League Score yet

The launch tables report no result for Neon on any of the five benchmarks the League scores. Our changelog records the check made today: LiveBench 2.0 — 37 configurations, read in full on November 19, 2026, unchanged, its top still the 84.9 tie between Claude Fable 5.2 and GPT-6.2 Astra — carries no Neon result, and neither does the BenchLM GPQA Diamond leaderboard, updated November 19, 2026. The headline 61 is the Artificial Analysis Intelligence Index, which the League does not score, so it cannot be converted into a League Score without inventing a mapping — and we do not invent mappings. The model therefore enters unranked at 0% coverage, exactly as our score coverage explainer describes for launch-day models.

Unranked is a statement about evidence, not about quality — the family’s own history is the proof. Gemini 4 Carbon entered unranked and recorded a LiveBench 83.5 two days later, which normalises to the 97.3 it holds today at 40% coverage; Gemini 4 Hydrogen did the same and entered at 85.0, as our Gemini 4 Hydrogen release analysis records, and Muse Ember entered the same way last week on the same headline index. Neon is one of 13 unranked models among the 67 we track, against 54 scored models in the AI model rankings — its first tracked score is the number that will place it.

What the unranked entry does not yet prove

None of that dims the release itself: a tier adapted from the family’s frontier model, priced flat with no introductory cliff of its own, between Hydrogen’s $30.00 workload and Carbon’s December $80.00, is a deliberate price position — on an unmeasured record. The score to watch is the first one: whichever tracked leaderboard publishes on Neon first will decide whether the middle tier is also the value tier. The model’s full record lives on its Gemini 4 Neon page, and the standings themselves live in the AI model rankings.