🔍 Read the full analysis: Gemini 4 Neon’s Modest Gains: Is The Lower Bill The Bigger Story? on ThorstenMeyerAI.com
Get hardware and tech essentials delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
Google released Gemini 4 Neon on November 18, describing it as a model that balances depth and efficiency. Artificial Analysis measured a one-point score gain over Gemini 4 Carbon at the highest setting, alongside a lower reported cost and shorter task time; the results vary by effort setting and test.
Google released Gemini 4 Neon on November 18, positioning it as a balance between the depth of Gemini 4 Carbon and the speed of Gemini 4 Hydrogen. Independent measurements from Artificial Analysis show Neon scoring 61 on its Intelligence Index at the xhigh setting, one point above Carbon at high, while costing less and completing the measured task faster; those results depend on model settings and benchmark workload.
On Artificial Analysis’s Intelligence Index, Neon at xhigh scored 61 points, compared with Carbon’s 60 at high. The index places Neon third among model families, behind GPT-6.2 Astra and Claude Fable 5.2. At standard list prices, the reported cost per index task was $6.44 for Neon at xhigh and $7.76 for Carbon at high, with task times of 21.5 and 37.1 minutes, respectively. Because those comparisons use different effort settings, they are not a like-for-like setting comparison.
At the same high setting, Neon scored 60 and cost $3.74 per index task. Moving Neon from high to xhigh added one index point, while the measured cost rose to $6.44 and task time to 21.5 minutes from 14.8. Artificial Analysis also reports lower standard token prices for Neon: $3 per million input tokens and $15 per million output tokens, versus Carbon’s standard $4 and $20. Carbon’s prices are discounted to $2 and $10 through November 30, according to the source material.
Results differ across tests. On the Artificial Analysis Coding Agent Index, Neon scored 58.6, close to Claude Fable 5.2 at 58.5 and Claude Opus 5.5 at 58.2; the reported task cost was $6.86 for Neon, compared with $15.47 and $12.61 for the two Claude models. Neon scored below Carbon on Terminal-Bench 4.0, 63.6% against 65.2%. Google’s own evaluation table marked Neon best in 16 of 19 rows, but compared it with Claude Opus 5.5 and GPT-6.2 Sol rather than the two models leading the independent index.
Gemini 4 Neon
Four models in seven weeks
Google’s Gemini 4 releases side by side, each at its best tested setting in the independent Artificial Analysis index. The best value in each row is outlined.
- Four models in seven weeksArgon on September 30, Carbon on October 28, Hydrogen on November 12, Neon on November 18. Neon is adapted from Carbon.
- A small step in the indexNeon at xhigh scores 61, one point above Carbon at high. It ranks third among model families, behind GPT-6.2 Astra and Claude Fable 5.2.
- A bigger step in cost and timeOne index task costs $6.44 instead of $7.76 and takes 21.5 instead of 37.1 minutes. Neon’s tokens are 25% cheaper than Carbon’s standard price.
- What to watchUntil November 30, Carbon still sells at a discounted $2/$10. Google’s own table compares Neon with Claude Opus 5.5 and GPT-6.2 Sol, not with the index leaders.
Source: Artificial Analysis, Gemini 4 Neon model page and article, accessed Nov 19, 2026 (Intelligence Index v5.0.1, cost and time per task at standard prices, release dates); Google, “Introducing Gemini 4 Neon”, Nov 18, 2026; Gemini 4 Neon model card. Output price in US$ per million tokens, standard rates. Redrawn by Thorsten Meyer AI.
Graphic: Powered by Thorsten Meyer AI
Cost and Speed Shape the Trade-Off
The release’s practical significance may lie less in a large jump in measured capability than in the balance between quality, price and completion time. For teams running repeated model tasks, a lower per-task bill or shorter runtime can matter even when a benchmark score changes only slightly. Artificial Analysis describes Neon as a small Intelligence Index gain over Carbon alongside a roughly 17% lower cost per task in its stated comparison.
That is not a universal saving estimate. The reported task costs reflect particular benchmark runs, model settings and standard prices; real costs can vary with prompt length, output volume, caching and workload. The source also says Neon uses about 15% more output tokens per task than Carbon, partly offsetting its lower token rates. Buyers should distinguish the published list price from Carbon’s temporary discount and test the settings their applications would actually use.
The model’s position among alternatives is also mixed. Its Coding Agent Index score is close to two Claude models, while the Intelligence Index ranks it behind GPT-6.2 Astra and Claude Fable 5.2. The available figures support a comparison of specific tests, not a claim that one model is best for every use.
Products Worth Considering

LG UltraGear 32GX870A-B 32-inch 4K OLED Dual-Mode Gaming Monitor

Logitech MX Keys S Combo Wireless Keyboard and Mouse, Graphite

Apple MacBook Air 13.6-inch with M5 Chip, 16GB Memory, 512GB SSD, Silver
As an affiliate, we earn on qualifying purchases.
Neon’s Place in Google’s Model Line
Neon is the fourth model Google has released in roughly seven weeks, following Gemini 4 Argon on September 30, Gemini 4 Carbon on October 28 and Gemini 4 Hydrogen on November 12. Google describes Carbon as suited to demanding, long-horizon projects and Hydrogen as intended for fast everyday execution, with Neon between them. Its model card says Neon is adapted from Carbon and directs readers to Carbon’s documentation for details about architecture, training data and hardware.
Google lists support for text, images, audio and video as inputs, with text output and a one-million-token context window. The default maximum output is 64,000 tokens, with up to one million available through background long decode. The listed knowledge cutoff is March 2026. Neon offers low, medium, high and xhigh effort settings, with high set as the default.
Google says Neon is available through the Gemini API in AI Studio, Gemini for business, and the Gemini app. Access through Google Antigravity and the Gemini app is listed for paid Pro and Ultra subscribers. A cyber-defense version is being made available to trusted defenders in Google’s Fairwind Program, with eligible members expected to receive access over the following few days.
““Frontier-level intelligence with fast end-to-end task completion.””
Benchmark Scope and Open Questions
The available comparisons do not settle how Neon performs across every real-world workload. Artificial Analysis’s reported figures are benchmark measurements, and task cost and speed can change with effort level, prompt, output length and pricing. The score comparison between Neon xhigh and Carbon high also uses different settings, while the same-setting Neon results show that higher effort comes with higher measured cost and longer runtime.
Google’s 19-row evaluation table does not include Claude Fable 5.2 or GPT-6.2 Astra, which lead the Artificial Analysis Intelligence Index. The supplied material does not establish why Google chose that comparison set, and the two sources’ individual tests should not be treated as interchangeable. Google also reports its lowest fabrication rate to date in internal Antigravity testing, but that is a separate measurement that outside readers cannot verify from the information provided.
It remains unclear how Neon will perform in broader independent evaluations, whether its benchmark cost advantage will hold for customer workloads, and how the temporary Carbon discount affects comparisons for users choosing between the two models before November 30.
Watch Independent Tests and Pricing
The next useful evidence will be additional independent testing across coding, reasoning and other production tasks, with matched model settings and clearly stated prices. Those comparisons can show whether Neon’s reported speed and cost differences persist outside the particular Artificial Analysis workloads cited here and how it fares against the models absent from Google’s table.
Users assessing the model can compare its low, medium, high and xhigh settings against their own quality, latency and budget needs. The source material gives no later benchmark schedule or further release milestone. Google’s stated near-term availability update is access to Gemini 4 Neon for cyber for eligible Fairwind Program members over the next few days; Carbon’s stated promotional pricing runs through November 30.
Key Questions
How much higher did Neon score than Carbon?
Artificial Analysis reported a score of 61 for Neon at xhigh and 60 for Carbon at high. That comparison uses different effort settings; Neon at high scored 60.
Is Neon cheaper than Carbon?
At standard list prices, Neon is listed at $3 per million input tokens and $15 per million output tokens, compared with Carbon’s $4 and $20. The source says Carbon is discounted to $2 and $10 through November 30, so the price comparison depends on when and how a customer buys.
Does Neon outperform Carbon on every benchmark?
No. In the cited independent results, Neon scored 63.6% on Terminal-Bench 4.0, below Carbon’s 65.2%. The outcomes vary by test and model setting.
Which models lead the Artificial Analysis Intelligence Index?
The source material identifies GPT-6.2 Astra and Claude Fable 5.2 as the two models ranked ahead of Neon among model families. Google’s cited comparison table does not include either model.
Where is Gemini 4 Neon available?
Google lists the Gemini API in AI Studio, Gemini for business and the Gemini app, with some app and Antigravity access tied to paid Pro and Ultra subscriptions. The cyber-defense version is being offered to eligible trusted defenders through the Fairwind Program.
Source: ThorstenMeyerAI.com
