what is gemini 4 neon? price, benchmarks and neon vs argon
gemini 4 neon is a google model announced on november 18, 2026 that sits between gemini 4 hydrogen and gemini 4 carbon. google built it for complex coding, analytical reasoning and production-grade enterprise workflows, and lists it on the paid gemini api, in google antigravity and for google ai pro and ultra subscribers in the gemini app.
its api price is in the table below. this guide follows google's launch post, the gemini api docs and the neon model card, as of november 2026.
download the ReMynd desktop app to search text you saw in a focused window, by app and date.
where does neon sit in the gemini 4 family?
neon is the middle option. google's choosing a model guide says neon has deeper reasoning than hydrogen and faster task completion than carbon, and its model card says neon is adapted from gemini 4 carbon.
| model | speed, per google | what google says to use it for |
|---|---|---|
| gemini 4 carbon | deliberate | your hardest problems: system architecture, debugging, science and security |
| gemini 4 neon | moderate | complex software engineering, dense knowledge work and science workflows |
| gemini 4 hydrogen | fast | the default for everyday coding and agentic workflows |
google's guide says carbon, neon and hydrogen share a 1m unified context window, long decode and the same built-in tools, and differ in thinking tokens per turn, speed and cost.
how do you use gemini 4 neon?
google's launch post lists where it is rolling out, and the gemini api docs give the model id gemini-4-neon.
| who you are | where google lists neon |
|---|---|
| developer | the gemini api paid tier in google ai studio and android studio; google antigravity for paid google ai pro or ultra subscribers |
| business | gemini for business, via the api or model drop-down, and google antigravity |
| consumer | the model drop-down in the gemini app, for google ai pro and ultra subscribers |
| cyber defender | eligible fairwind program members, who begin receiving access over the next few days |
to call it from the api, google's what's new in gemini 4 neon page gives these steps for apps already on another gemini 4 model:
- use a paid tier project, because neon is available on the paid tier only.
- set
modelto the neon model id above. google says no request schema changes are needed. - check the thinking level. neon defaults to
highand supportsxhigh, while carbon and argon default tomediumand cap athigh.
how much does gemini 4 neon cost?
google's gemini api pricing page lists these standard paid tier rates per 1m tokens. output includes thinking tokens, and google says every token the model processes is counted and billed.
| model | input | output | cached input |
|---|---|---|---|
| gemini 4 neon | $3.00 | $15.00 | $0.15 |
| gemini 4 argon or carbon, through november 30, 2026 | $2.00 | $10.00 | $0.10 |
| gemini 4 argon or carbon, from december 1, 2026 | $4.00 | $20.00 | $0.20 |
| gemini 4 hydrogen | $1.50 | $7.50 | $0.075 |
google's choosing guide says neon has no introductory price, and that hydrogen is on the free tier while carbon and neon aren't. the pricing page says argon is not on the free tier either. our ai model pricing comparison covers models from other companies.
how is neon different from gemini 4 argon?
google describes argon as its efficient gemini 4 frontier reasoning model, with faster responses and lower token spend per turn than carbon across complex coding, agentic, math and science tasks. in our read of google's api docs, the differences you can check against neon are price and thinking levels, both covered above.
google's neon model card adds one comparison. it says neon showed an increase in general capability compared with argon under google's material capability change assessment, but that this falls below the threshold that would require a full suite of critical capability evaluations, so google carries forward argon's results for those risk areas.
for argon itself, see is gemini 4 argon available and gemini 4 argon: what it is and who can use it.
what benchmarks does google report?
google's launch post gives these results for neon. google's neon evaluation methodology says all of the results are run with the highest thinking settings unless noted, and that all the results for non-gemini models are sourced from providers' self-reported numbers unless otherwise noted.
| benchmark | what google reports for neon |
|---|---|
| artificial analysis intelligence index | 61 at xhigh, #3 overall |
| vals index | 70.9%, #2 |
| artificial analysis coding agent index | 58.6, #1 overall |
| terminal-bench 4.1 | 65.2% |
| frontiercode 2.0 | 46.2%, at its default high setting |
| swe-sweep, in google antigravity | 39.0% |
google also says that at the default high thinking level, neon pairs a high artificial analysis intelligence index score with a 32% lower cost per task than carbon. these are the figures google reports, and we did not rerun any of them.
what are neon's limits?
google's model card names several. it says neon may show general limitations such as hallucinations, and that its knowledge cutoff is march 2026, though in some domains its knowledge may be less up to date. it also says the full safety and security report for neon will be available in the coming days.
on the api side, google's what's new page says neon does not support explicit caching, and that hard prompts can take several minutes, so it suggests background execution or http client timeouts of at least 15 minutes.
where does ReMynd fit?
ReMynd records your screen on a mac, runs ocr locally with apple vision, and keeps a searchable index on your computer. text comes from the focused window only, and video covers windows you have not excluded.
in our view, switching models costs you the context you gave the last one, and a new model does not fix that. ReMynd's agent window runs your own claude code or codex cli against that history, with read-only access, and connected agents send what they read to their own cloud. on a mac it needs macos 14 or later on an apple silicon mac.
ReMynd also runs on windows 11 (x64 or arm64).
for picking a coding model, see what is the best ai model for coding.
for google's interactive world model, see what is genie 4.
limits of this guide
- it follows google's launch post, the gemini api docs, the neon model card and google's evaluation methodology, as read on november 19, 2026. we did not test gemini 4 neon.
- the prices are google's gemini api list prices. gemini for business, google antigravity and gemini app plans were not checked for prices.
- google's launch post metadata shows it was edited the morning we read it, and model pages and prices change often, so check google's pages for the current ones.
download the ReMynd desktop app and search what you saw in a focused window, by app and date.