Google’s Gemini 4 Neon Table: The Benefits And Missing Details
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: Google’s Gemini 4 Neon Table: The Benefits And Missing Details on ThorstenMeyerAI.com

Before you orderOffer from Amazon

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
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

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 lower task costs and shorter runtimes; Google’s comparison table does not include the two models ahead of Neon in that independent index.

Google released Gemini 4 Neon on November 18, positioning it as a model between Gemini 4 Carbon and Gemini 4 Hydrogen for balancing capability with speed. In Artificial Analysis’s independent Intelligence Index, Neon scored 61 at its 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 and high setting.

The results point to a modest benchmark gain, with larger differences in measured cost and runtime. Artificial Analysis describes Neon as delivering “a small Intelligence gain over Gemini 4 Carbon at a ~17% lower Cost per Task.” Its figures put Neon’s score at 60 on the default high setting, matching Carbon’s score at high while costing $3.74 per index task. At xhigh, Neon’s score rises to 61 and cost to $6.44 per task. These are index-task estimates at standard list prices, not a promise of what every user will pay for a particular workload.

Artificial Analysis attributes the cost and speed changes to lower token prices and faster output, while noting Neon uses about 15% more output tokens per task than Carbon. Neon’s listed price is $3 per million input tokens and $15 per million output tokens, versus Carbon’s standard $4 and $20. Carbon is discounted to $2 and $10 through November 30, according to the source material, so the relative price depends on whether that promotion applies. Artificial Analysis measured about 110 output tokens per second for Neon, against about 55 for Carbon.

Google’s model card says Neon is adapted from Gemini 4 Carbon. It accepts text, images, audio and video, produces text, and has a one-million-token context window. Its default maximum output is 64,000 tokens, with up to one million available through background long decode. The model’s stated knowledge cutoff is March 2026. Google offers four effort settings—low, medium, high and xhigh—with high as the default. Availability listed in the source includes the Gemini API in AI Studio, business access, and certain paid Google AI Pro and Ultra offerings. A cyber-defense version is being made available to eligible members of Google’s Fairwind Program.

At a glance
reportWhen: Released November 18, 2026; benchmark d…
The developmentGoogle released Gemini 4 Neon, an adapted version of Gemini 4 Carbon that offers a small independent benchmark gain and lower measured task costs.
Crypto market snapshot
Fear & Greed Index
10/100 — Extreme Fear
Bitcoin BTC$63,649▼ 1.6%
Ethereum ETH$1,803▼ 0.4%
Tether USDT$0.9992▲ 0.0%
BNB BNB$589.03▼ 0.5%
USDC USDC$0.9996▲ 0.0%
XRP XRP$0.9801▼ 2.1%
Solana SOL$74.47▼ 0.4%
TRON TRX$0.3132▲ 1.5%
Live data · CoinGecko · alternative.me (24h change)

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.

ArgonCarbonHydro.Neon ReleasedSep 30Oct 28Nov 12Nov 18 Index50605461 Cost/task$4.97$7.76$3.16$6.44 Min./task29.737.115.321.5 Output $$10$20$7.50$15 Carbon $2/$10 discount ends Nov 30Neon: +1 point, $1.32 less per task
  1. 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.
  2. 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.
  3. 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.
  4. 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, Setting-Dependent Results

For developers and businesses running long or repeated model tasks, runtime and cost per task can matter as much as a small change in a benchmark score. Neon’s reported high-setting figures suggest users may get Carbon-like index performance for less at standard prices, while its faster measured output could shorten some runs. Those results offer a reason to test Neon against existing options, but do not establish savings for every application: actual costs depend on prompt length, output volume, caching, effort setting and workload.

The setting choice itself changes the trade-off. Artificial Analysis puts Neon at 60 points and $3.74 per task at high, compared with 61 points and $6.44 at xhigh. Medium scores 55 at $1.90, while low scores 44 at $0.75. A one-point increase from high to xhigh therefore comes with a substantial increase in the index’s estimated task cost. Teams considering deployment would need to evaluate whether the higher setting improves outcomes enough for their specific work.

Neon’s standing also depends on which competitors and tests are used. Artificial Analysis ranks it third among model families, behind GPT-6.2 Astra and Claude Fable 5.2. On the Coding Agent Index, Neon scores 58.6, close to Fable 5.2 at 58.5 and Claude Opus 5.5 at 58.2; the reported per-task costs are lower for Neon. These benchmarks can help frame comparisons, but they do not settle performance or reliability across all coding tasks.

Products Worth Considering

Logitech MX Mechanical Wireless Illuminated Keyboard, Tactile Quiet, Graphite

Logitech MX Mechanical Wireless Illuminated Keyboard, Tactile Quiet, Graphite

DJI Mini 4K Drone Combo with 4K Camera, 3-Axis Gimbal, and Two Batteries

DJI Mini 4K Drone Combo with 4K Camera, 3-Axis Gimbal, and Two Batteries

As an affiliate, we earn on qualifying purchases.

A Fourth Gemini 4 Model

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 material. Google describes Carbon as intended for demanding, long-horizon projects and Hydrogen for fast everyday execution, with Neon positioned between them. The model card’s description of Neon as adapted from Carbon is relevant to interpreting the launch: it is presented as a refinement of an existing model, not as a wholly separate model family.

Google’s launch post calls Neon “frontier-level intelligence with fast end-to-end task completion.” That is Google’s characterization, rather than an independent finding. Its model card refers readers to Carbon documentation for details on architecture, training data and hardware, leaving the supplied material without separate technical disclosures for Neon on those points.

Google’s published evaluation document contains 19 benchmark rows and marks Neon best in 16, according to the source. It also records rivals ahead in three: Claude Opus 5.5 on Terminal-bench 4.1 and FrontierCode 2.0, and GPT-6.2 Sol on Chartography. Artificial Analysis uses different tests and settings in its index. For example, its Terminal-Bench 4.0 result gives Neon at xhigh 63.6%, below Carbon at high at 65.2%. Results from different benchmark versions and settings should not be treated as direct equivalents.

“frontier-level intelligence with fast end-to-end task completion.”

— Google, in its Gemini 4 Neon launch post

Benchmark Coverage and Unverified Claims

Google’s comparison set leaves out two leading models in Artificial Analysis’s index: Claude Fable 5.2 and GPT-6.2 Astra. The source material says Google compares Neon with Claude Opus 5.5 and GPT-6.2 Sol instead. That omission does not invalidate Google’s published results, but it means its table does not show how Neon performed against those two independent-index leaders under the same reported evaluation.

Some measurements are also difficult to compare directly. Google’s table and Artificial Analysis use separate evaluations, and the Terminal-Bench versions cited are different. Google’s statement about Neon’s lowest fabrication rate is based on internal Antigravity testing, with no independently verifiable result supplied here. The available material also does not provide a separate account of Neon’s architecture, training data or hardware, beyond directing readers to Carbon’s documentation.

It remains unclear how the benchmark differences will translate to specific customer tasks, how widely the cyber-defense version will be available, and whether the listed prices or access terms will change. Benchmark scores and estimated task costs are not guarantees of real-world performance or spending.

Access, Pricing and Independent Tests

Neon is listed as available through the Gemini API in Google AI Studio and selected Google business and paid consumer offerings. Google says eligible members of its Fairwind Program will receive access to Gemini 4 Neon for cyber “over the next few days,” according to the source material. The dates and eligibility details for that rollout are not fully specified there.

The next useful evidence will be broader testing across comparable workloads and settings, alongside any further technical documentation from Google. Readers evaluating the model can compare costs at the effort level they expect to use and account for the stated Carbon discount through November 30. No single index score establishes which model is best for a particular task.

Key Questions

What is Gemini 4 Neon?

Google’s model card describes Gemini 4 Neon as adapted from Gemini 4 Carbon. Google positions it between Carbon and Hydrogen, with text, image, audio and video input and text output.

How much better did Neon score than Carbon?

Artificial Analysis scored Neon at 61 at xhigh, one point above Carbon at high, which scored 60. At the default high setting, Neon also scored 60, matching Carbon’s high-setting score.

Is Neon cheaper than Carbon?

Neon’s standard listed rates are $3 per million input tokens and $15 per million output tokens, below Carbon’s standard $4 and $20. The source says Carbon is discounted to $2 and $10 through November 30, so it may be cheaper during that promotion. Artificial Analysis estimated Neon’s index-task cost at $3.74 on high and $6.44 on xhigh.

Which leading models are missing from Google’s comparison table?

The source says 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 in Artificial Analysis’s index.

What remains unverified about Neon?

Google’s claim about its lowest fabrication rate comes from internal Antigravity testing, and the source says outsiders cannot check that measurement. Neon-specific details on architecture, training data and hardware are also not supplied separately in the cited model card material.

Source: ThorstenMeyerAI.com

Nothing in this article is financial or investment advice. Cryptocurrency and precious-metal investments carry significant risk — do your own research and consider a licensed advisor.
You May Also Like

Engineering Is Automated. Research Is the Residual.

Recent developments show AI can automate much of engineering, but research remains less automated. Experts see a shifting landscape in AI R&D capabilities.

Edge AI Vs Cloud AI: Where Does the Computing Happen?

Processing where AI computations occur—on local devices or remote servers—can dramatically affect performance and privacy; discover which suits your needs best.

World Model Readiness: Are You Ready for AI That Acts?

Assess your organization’s readiness for AI systems that predict and act, marking a shift from language models to world models capable of real-world action.

8 Best Guides To AI Automation Software For Workflow Building In 2027

A ThorstenMeyerAI.com roundup compares eight AI automation guides, from practical n8n projects to beginner instruction, self-hosting and business strategy.