🔍 Read the full analysis: The Trade-Off Behind Gemini 4 Neon: A Small Gain For A Lower Bill on ThorstenMeyerAI.com
Get smart everyday buys 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, positioning it as a balance of capability and efficiency. Artificial Analysis measured a one-point gain over Gemini 4 Carbon at the models’ highest reported settings, alongside lower cost and shorter task times; the result changes with the effort setting, and Google’s comparison table omits two models ahead of Neon on the independent index.
Google released Gemini 4 Neon on November 18, presenting it as a model that balances capability and efficiency. In Artificial Analysis measurements, Neon scored 61 at xhigh effort, one point above Gemini 4 Carbon at high effort, while an index task cost $6.44 and took 21.5 minutes, compared with $7.76 and 37.1 minutes for Carbon at its standard price. Those figures are not a like-for-like effort-setting comparison: at the shared high setting, Neon and Carbon scored 60, while Neon cost less.
Artificial Analysis describes the release as “a small Intelligence gain over Gemini 4 Carbon at a ~17% lower Cost per Task.” Its figures show that the apparent performance gain depends on the selected effort level. Neon at xhigh scored 61; at high, its default, it scored 60. Neon at high matched Carbon at high, also 60. At Neon’s high setting, the measured cost was $3.74 per index task and the task took 14.8 minutes. At medium, Neon scored 55 for $1.90 and 8.3 minutes; at low, it scored 44 for $0.75 and 3.7 minutes. Costs use standard list prices.
Neon’s listed rates are $3 per million input tokens and $15 per million output tokens, below Carbon’s standard $4 and $20 rates. Carbon was discounted to $2 and $10 through November 30, according to the source, meaning that temporary offer made Carbon cheaper on token prices during that period. Artificial Analysis also reported roughly 110 output tokens per second for Neon, versus about 55 for Carbon, while Neon used around 15% more output tokens per task, reducing some of the price advantage.
Google’s model card says Neon is adapted from Gemini 4 Carbon. It accepts text, images, audio and video, returns text, and has a one-million-token context window. The card lists a March 2026 knowledge cutoff and output of up to 64,000 tokens by default, with up to one million through background long decode. Google listed availability through its Gemini API in AI Studio, business products, and subscriptions to Google AI Pro and Ultra; access to a cyber-defense version is intended for eligible trusted defenders in its Fairwind Program.
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
Lower Costs Depend on Effort
For organizations running lengthy model tasks, Neon’s reported shorter completion times and lower standard-price task costs could matter more than its one-point top-setting score increase. But the index figures do not establish that every user or workload will see the same savings: real costs depend on prompt and output length, model settings, and the task being run. The lower score and price at medium or low effort also show that teams may need to choose between cost, speed and benchmark performance rather than treating Neon as a single fixed result.
Neon also performed near the top of the Artificial Analysis Coding Agent Index: Neon scored 58.6, Claude Fable 5.2 scored 58.5, and Claude Opus 5.5 scored 58.2. The reported run cost was $6.86 for Neon, compared with $15.47 for Fable and $12.61 for Opus. These are index results, not a guarantee of lower costs or equivalent performance across all coding work.
Products Worth Considering

FLEXISPOT Folding Walking Pad Treadmill

Canon SELPHY QX20 Compact Photo Printer – Dark Gray

FLEXISPOT 48 x 24-Inch Solid Rubberwood Electric Standing Desk, Black Frame
As an affiliate, we earn on qualifying purchases.
Google’s Rapid Gemini 4 Releases
Neon is Google’s fourth Gemini 4 model in about seven weeks, following Argon on September 30, Carbon on October 28, and Hydrogen on November 12, according to the source. Google described Carbon as suited to difficult, long-horizon projects and Hydrogen as intended for fast everyday execution, placing Neon between them. The model card’s note that Neon is adapted from Carbon provides a limited account of its relationship to that model; the source says the card refers readers to Carbon’s documentation for architecture, training data and hardware.
Google’s evaluation document lists results across 19 rows and marks Neon best in 16. It also records three exceptions: Claude Opus 5.5 leads Neon on Terminal-bench 4.1 and FrontierCode 2.0, while GPT-6.2 Sol leads on Chartography. The table compares Neon with those two models, but does not include Claude Fable 5.2 or GPT-6.2 Astra, which rank ahead of Neon on the Artificial Analysis Intelligence Index cited in the source. The two sets of results use different evaluations and comparison groups.
“frontier-level intelligence with fast end-to-end task completion.”
— Google, in its launch post
Benchmark Comparisons Have Limits
The available figures do not establish how Neon performs across every user workload, or whether its measured speed and cost differences persist under different prompts, settings and prices. The headline comparison between Neon xhigh and Carbon high uses different effort levels; the same-setting high comparison shows equal scores. Carbon’s stated discount through November 30 also means its temporary price differs from the standard-price basis used in the cost comparison.
Neon did not lead every independent test. On Terminal-Bench 4.0, it scored 63.6% at xhigh, below Carbon at high with 65.2%. On AA-Omniscience, Neon’s reported hallucination rate was 14.4%, compared with Carbon’s 11.4%. Google’s separate internal fabrication-rate claim cannot be independently checked from the information supplied, and the source does not provide enough detail to reconcile it with the independent result. The reasons for Google’s selected comparison set and the performance of omitted models on its 19-row table are also not established here.
Access and Testing Remain Key
Neon is available through the channels Google listed, including the Gemini API in AI Studio and eligible paid subscriptions. The source says eligible Fairwind Program members were expected to receive access to the cyber-defense version over the next few days; it does not confirm when that access was completed. The source also describes Carbon’s lower token prices as a discount ending November 30.
For users deciding whether to adopt Neon, the next useful evidence will be comparisons using the same effort settings, current prices and representative workloads. The supplied material does not report later benchmark updates or disclose additional independent evaluations. Until those are available, the clearest confirmed distinction is a small score gain at Neon’s highest setting, with lower reported standard-price task costs and faster runs in the cited index measurements—not a broad claim that Neon is better for every task.
Key Questions
What changed with Gemini 4 Neon?
Google released Neon on November 18, 2026, describing it as a balance of capability and efficiency. Artificial Analysis reported a one-point score gain over Carbon when comparing Neon at xhigh with Carbon at high, along with lower cost and shorter task time in its index figures.
Is Neon more capable than Carbon?
The cited results do not show a clear gain at the same high setting: both scored 60. Neon scored 61 at xhigh, but that compares against Carbon at high, so the effort settings differ. Neon also scored below Carbon on the cited Terminal-Bench 4.0 result.
How much does a Neon index task cost?
Artificial Analysis reported $3.74 per task at high effort and $6.44 at xhigh, using standard list prices. These are figures for its index task, not a fixed price for every real-world workload.
Does Google’s comparison table include the top independent-index models?
Not according to the source. Google’s 19-row table compares Neon with Claude Opus 5.5 and GPT-6.2 Sol, but does not include Claude Fable 5.2 or GPT-6.2 Astra, which rank ahead of Neon on the Artificial Analysis Intelligence Index cited there.
Source: ThorstenMeyerAI.com
