Gemini 4 Neon Compared: A Slight Edge At A Lower Price
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: Gemini 4 Neon Compared: A Slight Edge At A Lower Price on ThorstenMeyerAI.com

Age 18–24?Offer from Amazon

Prime made for students and young adults

  • Fast, free delivery for dorm and study essentials
  • Prime Video and Amazon Music included
  • Member-only deals
Try Prime for Young Adults Free trial for eligible 18–24 year olds
As an affiliate, we earn on qualifying purchases.

TL;DR

Google released Gemini 4 Neon on November 18, 2026. Artificial Analysis rates it one point above Gemini 4 Carbon at its highest effort setting, with lower per-task cost and shorter task times; the results vary by effort level and test. Google’s published comparison favors Neon in 16 of 19 rows but excludes two models that lead the independent index.

Google released Gemini 4 Neon on November 18, presenting it as a balance of capability and efficiency. In figures from independent evaluator Artificial Analysis accessed the next day, Neon scored 61 on the Intelligence Index at its xhigh setting, one point above Gemini 4 Carbon at high, while its measured index task cost was lower and its task time shorter. The comparison is not a universal win: results change by effort setting and benchmark, and Google’s own table does not include two models that lead the independent index.

Artificial Analysis reported a cost of $6.44 per index task for Neon at xhigh, compared with $7.76 for Carbon at high and standard list pricing. The measured times were 21.5 minutes for Neon and 37.1 minutes for Carbon. At the default high setting, Neon scored 60—the same score as Carbon in the comparison—and cost $3.74 per index task. These are benchmark results, not a guarantee of cost or performance for every user workload.

Neon’s standard API rates are $3 per million input tokens and $15 per million output tokens, against Carbon’s standard $4 and $20. Carbon was discounted to $2 and $10 through November 30, according to the source material, making the near-term price comparison different from the standard-rate comparison. Artificial Analysis attributes Neon’s lower measured task cost to lower token prices and faster output, while noting that Neon used about 15% more output tokens per task than Carbon.

Performance depends on the chosen effort level. Artificial Analysis lists Neon’s scores and costs as 61 and $6.44 at xhigh, 60 and $3.74 at high, 55 and $1.90 at medium, and 44 and $0.75 at low. On Terminal-Bench 4.0, Neon at xhigh scored 63.6%, below Carbon at high with 65.2%. The available figures therefore show a trade-off among score, cost and time rather than one setting that is best for every task.

At a glance
reportWhen: Released November 18, 2026; figures acc…
The developmentGoogle launched Gemini 4 Neon, a model adapted from Gemini 4 Carbon that posts a marginally higher independent index score at a lower measured cost per task.

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

Cost Gains Depend on Effort

For developers and businesses running repeated model tasks, cost per completed task and time to completion can matter as much as a headline benchmark score. The Artificial Analysis figures suggest Neon may offer a less expensive, faster route to results close to Carbon’s on the index, particularly at the default high setting. Buyers still need to test their own prompts and workflows: benchmark tasks may not reflect their actual token use, latency requirements or accuracy needs.

The coding comparison is close among several models. On Artificial Analysis’s Coding Agent Index, Neon scored 58.6, Claude Fable 5.2 scored 58.5 and Claude Opus 5.5 scored 58.2. The source reports a $6.86 task cost for Neon, against $15.47 for Fable 5.2 and $12.61 for Opus 5.5. Those figures may make Neon relevant for teams comparing coding agents, but a score difference of less than half a point does not establish that one model is better across coding work generally.

The release also highlights why model comparisons require care. Google’s 19-row evaluation table marks Neon best in 16 rows, but compares it with Claude Opus 5.5 and GPT-6.2 Sol—not Claude Fable 5.2 or GPT-6.2 Astra, which are ahead of Neon on the Artificial Analysis Intelligence Index. The tables answer different questions and use different evaluation setups; neither alone settles which model is the best fit.

De'Longhi Magnifica Duo Fully Automatic Espresso Machine with Milk Frother, Silver

De'Longhi Magnifica Duo Fully Automatic Espresso Machine with Milk Frother, Silver

KitchenAid Artisan Mini Tilt-Head Stand Mixer, 3.5 Quart, Aqua Sky Blue

KitchenAid Artisan Mini Tilt-Head Stand Mixer, 3.5 Quart, Aqua Sky Blue

Bose QuietComfort Ultra Wireless Noise Cancelling Earbuds, 2nd Gen, with 20W 2-Port Wall Charger

Bose QuietComfort Ultra Wireless Noise Cancelling Earbuds, 2nd Gen, with 20W 2-Port Wall Charger

As an affiliate, we earn on qualifying purchases.

Neon Builds on Gemini Carbon

Google describes Neon as sitting between Gemini 4 Carbon, intended for demanding long-horizon projects, and Gemini 4 Hydrogen, aimed at faster everyday execution. Its model card says Neon is adapted from Carbon and directs readers to Carbon documentation for details on architecture, training data and hardware. That limits what can be independently inferred from the launch announcement about how the models differ internally.

Neon is the fourth Gemini 4 model released in about seven weeks, following Argon on September 30, Carbon on October 28 and Hydrogen on November 12, according to the source. Google lists text, image, audio and video inputs, text output and a one-million-token context window. Its knowledge cutoff is March 2026; the model card lists output up to 64,000 tokens by default and up to one million through background long decode.

Google says Neon is available through the Gemini API in AI Studio and through several paid Google products, including Gemini for business and Google Antigravity for eligible Pro and Ultra subscribers. A cyber-defense version is being made available to trusted defenders in Google’s Fairwind Program, with access for eligible members expected over the next few days, according to the source material.

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

— Google, in its launch post

Benchmarks Leave Comparisons Open

The reported scores are benchmark measurements, not independent proof of performance across all uses. The source gives Artificial Analysis results accessed November 19, 2026, but does not provide enough detail here to reconcile every difference between its tests and Google’s evaluation table. Google’s table also compares Neon with a narrower set of rivals than the independent index, so its 16-of-19 result should not be read as a ranking against every leading model.

It is also unclear how representative the measured task costs are for a particular customer. Token volume, effort setting, caching, prompt design and output length can affect bills. The source says Neon produces about 110 output tokens per second versus Carbon’s roughly 55 on the cited model-page data, but does not establish that users will see the same speed in every deployment. Google’s internal fabrication-rate claim cannot be checked from the external figures provided.

Finally, the price comparison has a date-sensitive caveat: Carbon’s reported discount runs through November 30, 2026. The source material does not state what Carbon will cost afterward or whether Google will change Neon pricing. The launch information also does not provide a direct account of how Neon’s adapted status affects performance or behavior beyond the published test results.

Test Results Against Real Workloads

The next useful evidence will come from independent testing across more tasks and matched effort settings, along with comparisons that include GPT-6.2 Astra and Claude Fable 5.2. Such results would help clarify whether Neon’s cost advantage holds when models are evaluated on the same workloads, with comparable settings and price assumptions.

Teams considering Neon can compare the high and xhigh settings against their own accuracy, speed and budget needs, while checking the applicable API rates and product access. The source material gives no date for a further Gemini 4 release or for updated benchmark evaluations. Google’s Fairwind Program access for eligible cyber defenders was expected over the days following the November 18 launch.

Key Questions

How much higher did Neon score than Carbon?

Artificial Analysis reported 61 for Neon at xhigh and 60 for Carbon at high on its Intelligence Index. At high, Neon also scored 60. These figures compare specific effort settings, not necessarily identical configurations.

Is Gemini 4 Neon cheaper than Carbon?

At standard list prices, Neon is listed at $3 per million input tokens and $15 per million output tokens, versus Carbon at $4 and $20. Carbon was discounted to $2 and $10 through November 30, 2026, according to the source, so it was temporarily cheaper at those rates.

Does Neon lead every benchmark in the report?

No. The source reports that Neon at xhigh scored 63.6% on Terminal-Bench 4.0, below Carbon at high with 65.2%. Google’s table also records Claude Opus 5.5 and GPT-6.2 Sol ahead of Neon in three of its 19 rows.

Which competing models were not in Google’s comparison table?

The reported Google table compares Neon with Claude Opus 5.5 and GPT-6.2 Sol. It does not include Claude Fable 5.2 or GPT-6.2 Astra, the two models ahead of Neon on the Artificial Analysis Intelligence Index.

Where is Gemini 4 Neon available?

Google lists availability through the Gemini API in AI Studio and several Google products, including Gemini for business and Google Antigravity for eligible paid subscribers. A cyber-defense version is being offered to eligible trusted defenders through the Fairwind Program.

Source: ThorstenMeyerAI.com

You May Also Like

Chrome’s Eighth Zero-Day This Year: The XML Bug And Your Next Update

Google released Chrome 157 security updates for CVE-2026-125350, an XML use-after-free flaw it says is being exploited in the wild.

AI Alignment Sounds Abstract—But It Shapes What Machines Are Allowed to Do

Understanding AI alignment is crucial because it determines how machines act ethically, guiding their behavior in ways that directly impact human safety and values.

IdeaClyst: The Validation Council

IdeaClyst launches a new model council system for rigorous idea validation, using opposing AI models to stress-test ideas before roadmapping.

A Closer Look At Ring’s Flying Home Drone, Patrol Robot, And Privacy

Ring announced a home security drone and a business patrol robot, raising questions about connectivity, recording and video privacy.