🔍 Read the full analysis: Gemini 4 Neon, Reviewed: Modest Gains, Lower Costs, And Gaps In Google’s Table on ThorstenMeyerAI.com
Get movie nights 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 index gain over Carbon at the highest setting, alongside lower cost and shorter task times; its coding score is close to leading Claude models. Google’s comparison table does not include the two models ranked ahead of Neon by Artificial Analysis.
Google released Gemini 4 Neon on November 18, presenting it as a balance of capability and efficiency, while early independent measurements show a small score gain over Gemini 4 Carbon and lower cost and completion time on a benchmark task. The launch matters to developers weighing model performance against operating expense, but Google’s published comparison omits the two models ranked ahead of Neon on the independent index.
On Artificial Analysis’ Intelligence Index, Neon scored 61 at xhigh, compared with Carbon’s 60 at high. Artificial Analysis characterized the release as a small intelligence gain over Carbon alongside an approximately 17% lower cost per task. At standard list prices, the index task cost $6.44 with Neon and $7.76 with Carbon, and took 21.5 minutes rather than 37.1 minutes. These figures compare different effort settings, so they do not isolate the model change alone.
At the same high setting, Neon scored 60, matching Carbon’s high-setting score, while the reported cost per index task was $3.74 for Neon. Neon’s lower standard token rates contribute to the price difference: Google lists $3 per million input tokens and $15 per million output tokens, against Carbon’s standard $4 and $20. Carbon is discounted to $2 and $10 until November 30, according to the source report, so the standard-price comparison does not reflect that temporary offer.
The model is adapted from Gemini 4 Carbon, according to Google’s model card. It accepts text, images, audio and video, returns text, and has a one-million-token context window. Google lists four effort settings, with high as the default. The model is available through the Gemini API in AI Studio and several Google products; a cyber-defense version is being offered to eligible trusted defenders through Google’s 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 Shape Neon’s Appeal
Neon’s clearest reported advantage is not a large jump in benchmark intelligence, but the combination of lower standard prices and faster task completion. For teams running repeated model evaluations or agent workflows, time and cost per task can affect both budgets and how much work a system can complete. The reported figures give buyers a reason to test Neon, but they are benchmark results, not a guarantee that every production workload will be cheaper or faster.
Effort settings also change the trade-off. Artificial Analysis reports a cost of $3.74 per index task at high and $6.44 at xhigh, for a one-point score difference. Medium scored 55 at $1.90 per task, while low scored 44 at $0.75. Developers therefore have a practical choice between spending more for a marginal score gain and using a lower setting where speed or cost matters more. Results will depend on the task and the quality threshold a user needs.
In coding, the reported scores are close: Neon scored 58.6 on the Coding Agent Index, compared with 58.5 for Claude Fable 5.2 and 58.2 for Claude Opus 5.5. The associated task costs were $6.86, $15.47 and $12.61, respectively. That comparison suggests a potential cost advantage in this test, but the near-tied scores and benchmark-specific setup do not establish that the models perform equally across real software projects.
Popular right now

De’Longhi Rivelia Fully Automatic Espresso Machine with Grinder and Milk Frother, Beige

Apple AirPods Pro 3 Wireless Earbuds with Active Noise Cancellation

TCL 65-Inch R7D Series 4K UHD QLED Roku Smart TV
As an affiliate, we earn on qualifying purchases.
Google’s Rapid Gemini 4 Releases
Neon is the fourth Gemini 4 model Google has released in about seven weeks, following Argon on September 30, Carbon on October 28 and Hydrogen on November 12, according to the source report. Google describes Carbon as suited to difficult, long-horizon projects and Hydrogen as a fast option for everyday execution, positioning Neon between them.
Google’s launch post described Neon as delivering “frontier-level intelligence with fast end-to-end task completion.” Its model card adds that Neon is adapted from Carbon and directs readers to Carbon’s documentation for architecture, training data and hardware. The listed knowledge cutoff is March 2026; output is up to 64,000 tokens by default, with up to one million tokens available through background long decode.
Artificial Analysis’ index places Neon third among the model families cited in the source, behind GPT-6.2 Astra and Claude Fable 5.2. The comparison is not a universal ranking of every capability: individual benchmark results vary. For example, Neon scored 63.6% on Terminal-Bench 4.0 at xhigh, below Carbon at high, which scored 65.2%.
““frontier-level intelligence with fast end-to-end task completion.””
— Google, in its Gemini 4 Neon launch post
Benchmark Gaps and Limits
Google’s 19-row table compares Neon with Claude Opus 5.5 and GPT-6.2 Sol, but not with Claude Fable 5.2 or GPT-6.2 Astra, the two models ahead of Neon on Artificial Analysis’ Intelligence Index. Google marked Neon best in 16 rows; its table shows Claude Opus 5.5 ahead on Terminal-bench 4.1 and FrontierCode 2.0, and GPT-6.2 Sol ahead on Chartography. The source material does not explain why the two index leaders were left out of Google’s comparison set.
Benchmark results also differ by test and effort setting. On Artificial Analysis’ AA-Omniscience test, Neon had a reported hallucination rate of 14.4%, versus Carbon’s 11.4%; the cited rival figures were higher, but results from this test should not be treated as rates for all uses. Google’s claim of its lowest fabrication rate in internal Antigravity testing uses a different measure, and independent confirmation is not provided.
The available material does not establish how Neon’s benchmark performance translates to particular customer workloads, nor does it give enough detail to independently reproduce every comparison. The price figures also use standard list rates; Carbon’s temporary discount changes the cost comparison through November 30. Actual spending will depend on usage, settings and applicable pricing.
Availability and Buyer Testing
Neon is available through the Gemini API in Google AI Studio, Gemini for business, and Google Antigravity for paid AI Pro and Ultra subscribers; the Gemini app is also listed for Pro and Ultra subscribers. Google said eligible members of its Fairwind Program would receive access to the cyber-defense version over the following few days.
For users deciding whether to switch, the immediate next step is to test Neon on representative tasks at more than one effort setting and compare both output quality and full task costs. The source material provides no later benchmark release or announced review milestone. Further independent comparisons—including tests against the two models missing from Google’s table—would help clarify how the launch holds up beyond the currently reported indexes.
Key Questions
How much better did Gemini 4 Neon score than Carbon?
Artificial Analysis scored Neon at 61 at xhigh and Carbon at 60 at high on its Intelligence Index. At the same high setting, Neon scored 60, matching Carbon’s reported score at high.
Is Neon cheaper than Carbon?
At standard list rates, Google lists Neon at $3 per million input tokens and $15 per million output tokens, compared with Carbon at $4 and $20. Carbon is temporarily discounted to $2 and $10 through November 30, according to the source report, so the comparison changes during that offer.
Which models did Google’s comparison table leave out?
The table compared Neon with Claude Opus 5.5 and GPT-6.2 Sol. It did not include Claude Fable 5.2 or GPT-6.2 Astra, which Artificial Analysis ranked ahead of Neon on its Intelligence Index.
Where can users access Gemini 4 Neon?
Google lists Neon in the Gemini API in AI Studio, Gemini for business, Google Antigravity for paid AI Pro and Ultra subscribers, and the Gemini app for Pro and Ultra subscribers. Access to the cyber-defense version is for eligible trusted defenders through the Fairwind Program.
Source: ThorstenMeyerAI.com
