Model profile & tracker
Gemini 4 Neon: Price, Context and Specs (Updated)
Gemini 4 Neon (GA 18 November 2026): $3.00/$15.00 per million tokens, 1M context, four thinking levels. Specs, benchmarks, access, vs Carbon and Hydrogen.
Updated 22 November 2026. This page tracks Gemini 4 Neon — Google's reasoning model released on 18 November 2026 — and gets rewritten when Google changes the price, adds a free tier, or ships the cyber variant beyond its early program.
Neon is the middle model of the Gemini 4 family: "frontier-level intelligence with fast end-to-end task completion" for software engineering, knowledge work and science. It carries a 1M-token context window, four thinking levels up to xhigh, and a list price of $3.00 input / $15.00 output per million tokens, thinking included. It ranks #3 on the Artificial Analysis Intelligence Index at 61 — one point above Carbon — and #2 on the Vals Index.
The specs, as published
| Spec | Value |
|---|---|
| Model ID | gemini-4-neon |
| Released | 18 November 2026 (GA, paid tier) |
| Context window | 1,048,576 tokens, input and output combined |
| Output cap | 65,536 tokens by default, up to 1,048,576 with long decode |
| Thinking levels | low, medium, high (default), xhigh |
| Input price | $3.00 / 1M tokens (Batch and Flex: $1.50) |
| Output price | $15.00 / 1M tokens, thinking included (Batch and Flex: $7.50) |
| Cached input | $0.15 / 1M tokens — implicit caching only |
| Priority tier | $5.40 / $27.00 per 1M tokens |
| Free tier | Not available |
| Knowledge cutoff | March 2026 |
| Inputs | Text, images, video, audio, PDFs |
| Not supported at launch | Image and audio generation, explicit caching, Gemini Live API |
The caching note matters more than it looks. Implicit caching is automatic — you cannot mark blocks yourself, so whether your workload hits the 95% discount depends on Google's cache, not your prompt design.
The cost story Google published
Google's headline claim is 32% lower cost per task than Gemini 4 Carbon at the default high thinking level. The mechanism is positioning, not a price cut: Neon thinks less for a similar answer, so the bill shrinks even at comparable per-token rates.
| Model | Input / output per 1M | AA Index |
|---|---|---|
| Gemini 4 Carbon (promo to 30 Nov) | $2.00 / $10.00 | 60 |
| Gemini 4 Neon | $3.00 / $15.00 | 61 |
| Gemini 4 Carbon (from 1 Dec) | $4.00 / $20.00 | 60 |
Through 30 November, Carbon's introductory rates undercut Neon per token. When they double on 1 December, Neon becomes the cheaper way to buy the family's top score — 25% below Carbon's list price. The index figures are Artificial Analysis's, not Google's; the "61 vs 60" gap is within the noise of any one evaluation, so treat it as "same tier", not "Neon wins".
How to access it
Three routes, all live since launch day: the Gemini API on the paid tier at the prices above, the Gemini app on Pro and Ultra subscriptions, and Google Antigravity. A "Neon for cyber" variant is rolling out to Fairwind Program members. There is no free tier — that remains Hydrogen's job.
Neon vs the rest of the family
vs Gemini 4 Hydrogen ($1.50 / $7.50): Hydrogen is half the price with the same 1M window, a free API tier and the family's top speed rating — 5/5 on Google's own slide to Neon's 3.5/5 — but tops out at AA Index 54. Hydrogen is for volume and latency; Neon is for problems that need the extra thinking. Details on the Hydrogen page.
vs Gemini 4 Carbon: same tier of intelligence by every published index. Until 30 November, Carbon is cheaper per token on promo rates; after that, Neon is the default buy for the family's top score unless you specifically need Carbon's longest-horizon deep research.
vs Qwen4-Max ($1.50 / $4.50): Alibaba's flagship is half Neon's input price and less than a third of its output price for a comparable 1M window, at the cost of fewer input modalities (text and image vs five). If your stack is already on Google, Neon keeps you in AI Studio, Antigravity and the Gemini app; if cost per token rules, Qwen4-Max wins on bill.
FAQ
When was Gemini 4 Neon released? 18 November 2026, generally available on the paid tier the same day — no preview period.
Is Gemini 4 Neon on the free tier? No. The paid Gemini API is the only API route; Hydrogen is the family's free-tier model.
Is Neon better than Carbon? Neon scores 61 on the AA Intelligence Index at xhigh to Carbon's 60 — effectively the same tier. The real difference is price after 1 December and Carbon's focus on the hardest long-horizon research.
What is the context window? 1M tokens shared between input and output, with 65,536 output tokens by default and up to the full window via long decode.
Should you switch to it? If you are on Carbon at default settings after 1 December, yes on cost — Google's own 32%-per-task claim and the 25% list-price gap point the same way. If you are on Hydrogen, only switch for problems its AA Index 54 can't crack. Run your own evaluation against the comparison table before moving anything production.