Tech

News

Updated

Google releases Gemini 4 Neon, a model between Gemini 4 Hydrogen and Carbon

Key facts

Neon costs $3 per million input tokens and $15 per million output tokens on the paid Gemini API, and Pro and Ultra subscribers can pick it in the Gemini app.

On Nov. 18, 2026, Google released Gemini 4 Neon, a model it says brings "frontier-level intelligence with fast end-to-end task completion across complex software engineering, knowledge work, and science workflows," according to Google's announcement.

What happened

Koray Kavukcuoglu, senior vice president of Google DeepMind and Google's chief AI architect, announced the model on Google's blog. The post describes where Neon sits in the Gemini 4 family: Carbon "tackles your hardest, long-horizon projects," Hydrogen "handles fast everyday execution," and Neon "balances depth and efficiency."

Google's API guide gives the model ID as gemini-4-neon and says it sits between Gemini 4 Hydrogen and Gemini 4 Carbon for high-frequency complex tasks. The guide tells developers to use Carbon for the hardest architecture, debugging, science and security problems, to keep Hydrogen as the default for everyday coding and agentic work, and to use Gemini 3.5 Flash-Lite for high-volume, low-latency tasks.

The announcement also says Neon can act as an orchestrator that directs parallel Hydrogen subagents in Google Antigravity, and that Tasks in Gemini can now run on Neon.

Where it is available

Google lists these places in its announcement, which says the model begins rolling out on Nov. 18.

WhereWho gets it
Gemini appGoogle AI Pro and Ultra subscribers, from the model drop down
Gemini API in Google AI Studio and Android StudioDevelopers on the paid tier
Google AI StudioAll Pro and Ultra subscribers
Google AntigravityPaid Google AI Pro and Ultra subscribers, and enterprises
Gemini for businessThrough the API or the model drop down
Gemini 4 Neon for cyberEligible Fairwind Program members, over the next few days

The API guide says Neon is not on the free tier of the Gemini API, and that the Antigravity agent in Gemini Managed Agents supports it, on its own or paired with Hydrogen.

Price compared with Hydrogen and Carbon

Each price below is the paid tier price per 1 million tokens on Google's own guide for that model. Neon and Hydrogen use flat pricing across the full 1 million-token context window, according to the Neon guide and the Hydrogen guide.

ModelInputOutput
Gemini 4 Hydrogen$1.50$7.50
Gemini 4 Neon$3.00$15.00
Gemini 4 Carbon, through Nov. 30, 2026$2.00$10.00
Gemini 4 Carbon, after Nov. 30, 2026$4.00$20.00

The Carbon rows come from the Carbon guide, which lists an introductory price through Nov. 30, 2026. The Neon guide says implicit caching cuts the price of cached input tokens by 95%, to $0.15 per million, and that thinking tokens are billed as output tokens. Hydrogen has a free tier and Neon does not.

Google's benchmark figures

Google published these scores for Neon in its announcement. They are Google's figures and rankings as Google states them.

BenchmarkGoogle's figure for Neon
Artificial Analysis Intelligence Index, xhigh61, ranked #3 overall
Artificial Analysis Coding Agent Index58.6, ranked #1 overall
Vals Index70.9%, ranked #2
Terminal-Bench 4.165.2%
FrontierCode 2.0, high46.2%
Arena Agent Arena Index1336, ranked #2

The same post says Neon at its default high thinking level has "a 32% lower cost per task than Carbon," and that at xhigh it costs "36%+ less per task than competing models at their highest setting." On safety, Google says Neon took shortcuts on 2.45% of tasks on its coding cheating evaluation, which it calls the lowest rate of any Gemini 4 model, and points to the Gemini 4 Neon model card for details.

What changed for developers

Neon's API guide lists what changes when moving to it.

TopicWhat the guide says
From Carbon, Argon or HydrogenNo request schema changes; set the model to gemini-4-neon on a paid tier project
Default thinking levelNeon defaults to high and supports xhigh, like Hydrogen; Carbon and Argon default to medium and cap at high
From Gemini 3.8 Flash and earlierOne shared 1,048,576-token window replaces separate 1M input and 64K output budgets
Caching and samplingNo explicit caching; remove temperature, top_p and top_k

A single request defaults to and is capped at 65,536 tokens, and longer responses up to 1 million tokens need background execution or client-side continuation, according to the guide. It supports four thinking levels: low, medium, high and xhigh.

What comes next

Google says Gemini 4 Neon for cyber is "rolling out to trusted defenders in the coming days as part of the Fairwind Program," and that eligible members will begin receiving access over the next few days, according to its announcement. Carbon's introductory API price runs through Nov. 30, 2026, after which the Carbon guide lists $4 per million input tokens and $20 per million output tokens.

Background

Neon arrives six days after Gemini 4 Hydrogen, which Google released on Nov. 12, 2026, as the upgrade from Gemini 3.8 Flash and whose paid tier price per token is half of Neon's, according to the Hydrogen guide.

Have another question?