TL;DR
Google released Gemini 4 Neon on November 18. Artificial Analysis scores it one point above Gemini 4 Carbon at the tested top settings, while its measured index task costs less and takes less time. Carbon remains discounted through November 30, and independent rankings place two other model families ahead of Neon.
Google released Gemini 4 Neon on November 18, presenting it as a model that balances capability with speed. In Artificial Analysis’s independent Intelligence Index, Neon scored 61 at xhigh, one point above Gemini 4 Carbon’s score of 60 at high; the measured index task cost $6.44 and took 21.5 minutes, compared with $7.76 and 37.1 minutes for Carbon at its standard price.
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
The comparison uses each model’s best tested setting, so it does not hold effort level constant: Neon was tested at xhigh and Carbon at high. At the same high setting, Artificial Analysis reports that Neon matches Carbon’s score for $3.74 per task. The index score is a benchmark result, while the task cost and time refer to Artificial Analysis’s evaluation; they do not establish what every user will pay or how long every real-world job will take.
Neon’s standard token prices are $3 per million input tokens and $15 per million output tokens. Carbon’s standard rates are $4 and $20, but Google’s temporary Carbon discount lowers them to $2 and $10 through November 30. That means Carbon is cheaper per token during the promotion, even though Neon’s measured index task cost is lower than Carbon’s standard-price task cost.
Artificial Analysis puts Neon third among model families on its Intelligence Index, behind GPT-6.2 Astra and Claude Fable 5.2. On its Coding Agent Index, Neon scores 58.6, close to Claude Fable 5.2 at 58.5 and Claude Opus 5.5 at 58.2. The reported coding-task costs are $6.86 for Neon, $15.47 for Fable 5.2 and $12.61 for Opus 5.5. These figures are specific to the index evaluations.

Lower Costs, Close Coding Scores
For organizations running repeated model evaluations or coding-agent workloads, cost per completed task can matter as much as a small change in a benchmark score. Artificial Analysis’s measurements suggest Neon may reduce spending and waiting time on its index task relative to Carbon at standard prices, while keeping the Intelligence Index score close. The coding results also put Neon near the two Claude models tested, at a lower reported cost per benchmark task.
Those comparisons need qualification. Carbon’s promotion runs through November 30, so buyers choosing between the two models should compare current discounted rates with their own workloads. The index task is not a universal unit of work, and a model’s token rates, output speed and token usage can combine differently across prompts. Neon’s 61-point result is also below the scores of the two model families that lead the independent index.
Products Worth Considering

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

Apple MacBook Air 15-inch Laptop with M5 Chip, 16GB Memory, 512GB SSD, Midnight

Samsung Galaxy Tab A11+ Tablet (6GB RAM, 128GB Storage, Gray)
As an affiliate, we earn on qualifying purchases.
Four Gemini Releases Since September
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. Artificial Analysis’s table records scores of 50, 60, 54 and 61 respectively at each model’s best tested setting. Its measured cost and time per index task vary across that sequence, so the releases do not form a simple steady progression in every measure.
Google describes Carbon as intended for difficult, long-horizon projects, Hydrogen for fast everyday execution and Neon as a balance of depth and efficiency. The Neon model card says the model is adapted from Gemini 4 Carbon and points to Carbon’s documentation for architecture, training data and hardware details. Google lists text, image, audio and video input, text output and a one-million-token context window. Its stated knowledge cutoff is March 2026.
Google’s launch comparison reports Neon ahead in 16 of 19 rows against Claude Opus 5.5 and GPT-6.2 Sol. That vendor comparison does not include Claude Fable 5.2 or GPT-6.2 Astra, the two model families ahead of Neon in Artificial Analysis’s Intelligence Index. Results from those different comparisons should not be treated as a single shared ranking.
“frontier-level intelligence with fast end-to-end task completion”
— Google, in its Gemini 4 Neon launch post
How Neon Performs in Use
The supplied results do not establish how Neon performs across every customer workload, or whether the measured task cost and time will carry over to production use. Benchmark scores and index task costs depend on the evaluation’s prompts, settings and token use. The comparison also tests Neon at xhigh against Carbon at high for the headline scores, which limits what can be inferred about the score difference at matching effort levels.
Google’s comparison omits the two model families that lead Artificial Analysis’s index, while the independent coding scores cover only a small set of models. The source material does not provide a like-for-like result across all leading systems on the same broader set of tasks. It also does not quantify how Neon’s roughly 15% higher output-token use affects costs across different workloads.
Carbon Discount Ends November 30
The immediate pricing milestone is November 30, when the stated Carbon discount ends. Buyers comparing the models can reassess the price difference after that date, while accounting for the fact that actual costs depend on token use and workload. Google lists Neon as available through the Gemini API in AI Studio, Gemini for business, and specified paid subscriptions in Google Antigravity and the Gemini app.
Google also says Gemini 4 Neon for cyber will reach eligible trusted defenders in its Fairwind Program over the next few days. The supplied material gives no further deployment detail. More independent comparisons, especially at matching effort settings and against leading models on the same tasks, would help clarify how far Neon’s cost and speed results extend beyond the reported indexes.
Key Questions
How much higher did Neon score than Carbon?
Artificial Analysis scored Neon at 61 at xhigh and Carbon at 60 at high. Because the tested effort settings differ, the one-point gap is not a same-setting comparison. At high, the source says Neon matches Carbon’s score.
Is Neon cheaper than Carbon?
Neon’s standard rates are $3 per million input tokens and $15 per million output tokens, compared with Carbon’s standard $4 and $20. Carbon is temporarily discounted to $2 and $10 through November 30, making its promotional token rates lower during that period.
Which models rank ahead of Neon?
Artificial Analysis ranks GPT-6.2 Astra and Claude Fable 5.2 ahead of Neon among model families on its Intelligence Index. Google’s launch table compares Neon with Claude Opus 5.5 and GPT-6.2 Sol instead.
When was Gemini 4 Neon released?
Google released Gemini 4 Neon on November 18, 2026. The source material was accessed the following day.
Source: Thorsten Meyer AI
