🔍 Read the full analysis: Can Gemini 4 Neon Lower Costs Without Giving Up Much Performance? on ThorstenMeyerAI.com
Get the latest gadgets 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 more efficient model adapted from Gemini 4 Carbon. Artificial Analysis measured a one-point score gain at Neon’s xhigh setting, alongside lower per-task costs and shorter task times; results vary by effort setting and benchmark.
Google released Gemini 4 Neon on November 18, presenting it as a more efficient option between Gemini 4 Carbon and Gemini 4 Hydrogen. In measurements published by Artificial Analysis, Neon scored 61 on its Intelligence Index at the xhigh setting, one point above Carbon at high, 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.
The comparison does not show a like-for-like effort setting: Neon’s 61-point result uses xhigh, while Carbon’s 60-point result uses high. At the same high setting, Neon also scores 60, according to Artificial Analysis, for $3.74 per index task. The evaluator’s figures therefore point to a stronger cost and speed case than a large improvement in measured intelligence.
Google lists Neon at $3 per million input tokens and $15 per million output tokens, below Carbon’s standard rates of $4 and $20. However, Carbon is discounted to $2 and $10 through November 30, according to the source report. Neon’s price advantage over Carbon’s regular list price does not apply during that promotion. Artificial Analysis also reports that Neon produces about 110 output tokens per second, roughly twice Carbon’s rate, while using about 15% more output tokens per task.
Performance changes with the chosen effort level. Artificial Analysis gives Neon scores of 60 at high, 55 at medium and 44 at low, with respective task costs of $3.74, $1.90 and $0.75. Moving from high to xhigh raises the score by one point and the task cost to $6.44. Neon also does not lead every test: on Terminal-Bench 4.0, its xhigh score was 63.6%, below Carbon at high, which scored 65.2%.
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
The Cost of Each Performance Point
For teams paying to run lengthy reasoning or coding tasks, cost and turnaround time can matter as much as a top-line benchmark score. The reported index results suggest Neon may offer a lower-cost route to performance close to Carbon’s, particularly at the high setting. That could affect the economics of repeated model use, but the figures are benchmark-based estimates, not a guarantee of savings on every workload.
Neon’s coding results offer another point of comparison. 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. Artificial Analysis puts Neon’s task cost at $6.86, compared with $15.47 for Fable and $12.61 for Opus. Those costs and scores describe the evaluator’s test conditions; buyers should not treat them as direct forecasts for their own software projects.
Products Worth Considering

Amazon Kindle Scribe (16GB) with Premium Pen, Tungsten

Bose QuietComfort Wireless Noise Cancelling Headphones – Twilight Blue Limited Edition

Apple MacBook Air 15-inch (2026) with M5 chip, 16GB memory, 512GB SSD — Midnight
As an affiliate, we earn on qualifying purchases.
A Model Adapted From Carbon
Google describes Neon as balancing depth and efficiency between Carbon, intended for harder, longer projects, and Hydrogen, aimed at faster everyday execution. Its model card says Neon is adapted from Carbon and points readers to Carbon documentation for architecture, training data and hardware details. Google released Neon about seven weeks after Carbon on October 28, following Argon on September 30 and Hydrogen on November 12.
The model accepts text, images, audio and video, returns text, and has a one-million-token context window. Its default output limit is 64,000 tokens, with up to one million available through background long decode. Google gives a March 2026 knowledge cutoff and four effort settings, with high as the default. Availability includes the Gemini API in AI Studio, business offerings and specified paid Google AI Pro and Ultra services. A cyber-defense version is being made available to eligible members of Google’s Fairwind Program.
Google’s evaluation document reports Neon ahead in 16 of 19 rows against Claude Opus 5.5 and GPT-6.2 Sol. Those results are Google’s published comparisons, not the same as the independent index. The table does not include Claude Fable 5.2 or GPT-6.2 Astra, which the source report identifies as the two models above Neon in Artificial Analysis’s family rankings.
“While Carbon tackles your hardest, long-horizon projects and Hydrogen handles fast everyday execution, Neon balances depth and efficiency.”
Benchmark Gaps and Open Questions
The available results do not establish how Neon will perform across every customer workload. Benchmark scores depend on the test and effort setting, and the xhigh-versus-high comparison behind the headline one-point gain is not a matched-setting comparison. The source material also does not provide details that would let readers independently reproduce all reported task costs or translate them directly into their own usage bills.
Google’s internal fabrication-rate claim uses a different measure from Artificial Analysis’s AA-Omniscience result, so the two should not be treated as comparable. On that independent test, Neon’s reported hallucination rate was 14.4%, versus 11.4% for Carbon at the tested settings. Google’s benchmark table also leaves out two models that rank ahead of Neon in the independent family standings. The extent of that gap across Google’s full set of tests is not shown in the supplied material.
Testing Neon Against Real Workloads
The next useful evidence will come from matched-setting comparisons and customer use beyond benchmark suites: the same tasks, effort levels and pricing assumptions across Neon, Carbon and competing models. Buyers can also test whether Neon’s faster token generation offsets its greater output-token use in their own applications. Carbon’s stated discount ends November 30, which will change the price comparison after that date if the promotion is not extended.
Google says eligible Fairwind Program members will receive access to Gemini 4 Neon for cyber over the next few days. The source material does not specify when broader results, additional independent evaluations or updated comparisons against the models omitted from Google’s table will be available.
Key Questions
Is Gemini 4 Neon smarter than Gemini 4 Carbon?
Artificial Analysis scored Neon at 61 on xhigh and Carbon at 60 on high. At the same high setting, Neon also scored 60, so the reported one-point lead compares different effort levels and does not show a gain at matched settings.
How much does a Neon index task cost?
Artificial Analysis reports $6.44 per task at xhigh and $3.74 at high, using standard list prices. Its estimates are for index tasks; costs for a particular application can differ.
Is Neon always cheaper than Carbon?
Neon’s listed rates are below Carbon’s standard rates, but Carbon is discounted through November 30, 2026, at $2 per million input tokens and $10 per million output tokens. The source report does not establish what Carbon will cost after the promotion.
Where can people access Gemini 4 Neon?
Google lists access through the Gemini API in Google AI Studio, business offerings and specified paid Google AI Pro and Ultra services, including the Gemini app and Google Antigravity. The cyber-defense version is for eligible Fairwind Program members.
Source: ThorstenMeyerAI.com
