Google released Gemini 4 Neon on Wednesday, November 18, the fourth model in its Gemini 4 family in less than a month. It costs $3 per million input tokens and $15 per million output tokens, with a 1 million-token context window. Google positions it between Carbon, its heavyweight for long projects, and Hydrogen, its fast everyday model.
“While Carbon tackles your hardest, long-horizon projects and Hydrogen handles fast everyday execution, Neon balances depth and efficiency,” Google said in its announcement. The pitch is plain: close to Carbon’s brains, without Carbon’s bill.
Koray Kavukcuoglu, Google DeepMind’s chief AI architect, announced Neon in a blog post on November 18. It is rolling out now in the Gemini API, Google AI Studio, Gemini for Business and Google Antigravity, plus the Gemini app for paying Pro and Ultra subscribers. Free users don’t get it, and there is no free API tier.
Google shipped four flagship models in nine days.
Where Gemini 4 Neon fits in the Gemini 4 lineup
Neon is the fourth Gemini 4 model since November 10, when Google made Argon and Carbon broadly available. Hydrogen followed on November 12. Each one fills a different slot: Argon and Carbon take on the hardest work, Hydrogen handles quick everyday tasks, and Neon covers the middle ground where most real coding and analysis actually happens.
Neon is adapted from Carbon, according to its model card, and it shows. On the Artificial Analysis Intelligence Index, Neon scores 61 at its highest thinking level, ranking third overall. It ranks second on the Vals Index with 70.9%, and second on the Agent Arena Index with 1336 across more than 4,200 real-world agent sessions.
Four thinking levels let developers dial the reasoning depth up or down.
What Gemini 4 Neon costs
Neon’s standard API rates are $3 per million input tokens, $0.15 per million cached input tokens and $15 per million output tokens. Cached input gets a 95% discount, which matters for long coding sessions where the same files get read again and again. At the default high thinking level, Google says Neon costs 32% less per completed task than Carbon.
That pricing lands in a crowded middle. Hydrogen costs $1.50 per million input tokens and $7.50 per million output tokens, while Carbon’s introductory $2 and $10 rates rise to $4 and $20 on December 1. From that date, Neon will cost less than Carbon on input and output alike, which makes the mid-tier model the cheaper option for most heavy work.
Against rivals, Neon undercuts OpenAI’s GPT-6.2 Astra at $8 per million input tokens and $40 per million output tokens, though OpenAI’s cheaper Sol sits at $2 and $10 and Anthropic’s Claude Opus 5.5 matches Astra. Google is selling cost per finished task, and Neon is its entry in that fight.
Gemini 4 Neon benchmarks
Google’s benchmark table puts Neon ahead of Claude Opus 5.5 and GPT-6.2 Sol on most of the tests it lists. Neon leads on 16 of 19 comparisons, including the Artificial Analysis Intelligence Index, where it scores 61 against 59 for Opus 5.5 and 57 for GPT-6.2 Sol.
Coding is where Neon makes its case. It ranks first on the Artificial Analysis Coding Agent Index with a score of 58.6, and scores 65.4% on CursorBench 4.1, beating Carbon’s 58.1% and Hydrogen’s 57.3% in live IDE editing. On Terminal-Bench 4.1 it scores 65.2%, a hair behind Carbon’s 65.8%. It also tops Zapier’s AutomationBench at 56.3% and Andon Labs’ Vending-Bench 2 with $16,843.
Science holds up too: Neon scores 70.5% on Terminal-Bench Science, 84.0% on RiemannBench and 71.3% on HLE-Diamond, all best in class according to Google, plus 73.5% on Chartography for reading dense charts.
One caution applies to all of it: Google published the comparison table itself, and independent tests of Neon have not landed yet. Benchmarks from a vendor’s own harness are a starting point, not a verdict. Early users and outside labs will tell the real story over the next few weeks.
Who can use Gemini 4 Neon
Developers can reach Neon through the Gemini API in Google AI Studio and through Gemini for Business, by API or model dropdown. In the Gemini app it is limited to Pro and Ultra subscribers, and in Google Antigravity it is open to Pro and Ultra subscribers as well. GitHub’s changelog also lists Neon as generally available in Copilot Pro+, Max, Business and Enterprise plans.
Google has not said whether Neon will reach free Gemini users or AI Mode in Search, where Carbon and Hydrogen already sit. The model card lists a knowledge cutoff of March 2026 and warns that Neon can produce fabricated information, though Google says it recorded its lowest fabrication rate yet in internal Antigravity testing.
Why this matters
Four flagship models in nine days is not a normal release pace, even for Google. Neon fills the slot that matters most: the one where teams run agents all day, every day.
Price is now the battleground. OpenAI cut Astra’s price 20% when it launched GPT-6.2 this week, and Anthropic priced its record IPO the same day Neon arrived. Every lab is chasing the same customer, the developer who needs a model that finishes a coding task without burning the budget.
Neon answers that with a simple trade: most of Carbon’s ability, at a price that drops below Carbon’s on December 1, with coding scores that beat Carbon where it counts. If independent tests back up Google’s table, Neon becomes the default pick for agent-driven coding.
FAQ
What is Google Gemini 4 Neon?
It is Google’s newest AI model, launched November 18, 2026, for complex coding, knowledge work and science. It sits between Gemini 4 Carbon and Gemini 4 Hydrogen in the Gemini 4 family.
How much does Gemini 4 Neon cost?
Neon costs $3 per million input tokens and $15 per million output tokens through the API. Cached input costs $0.15 per million tokens, a 95% discount.
How does Gemini 4 Neon compare to Carbon?
Google says Neon costs 32% less per completed task than Carbon at the default high thinking level, while scoring close to Carbon on most benchmarks. Neon ranks first on the Artificial Analysis Coding Agent Index, ahead of Carbon.
Is Gemini 4 Neon free?
No. Neon is available to paid Google AI Pro and Ultra subscribers and through the paid Gemini API. There is no free tier, and Google has not said when or if free users will get it.
Sources: Google blog, The Verge, Tbreak, Gadgets 360, Moneycontrol.
