Gemini 4 Neon vs Carbon: is the faster model good enough?

AIHubMix11 min read

Gemini 4 Neon isn't a new model line. Google's model card says so in one line: "Gemini 4 Neon is adapted from Gemini 4 Carbon." On Artificial Analysis, Neon at its default high setting scores 59.6 on the Intelligence Index against 60.0 for Carbon High, and finishes an average task in 14.8 minutes against 36.6.

The price comparison depends on the date. Today Neon costs 50% more per token than Carbon, which is still on introductory rates. On December 1 Carbon's rates double, and Neon's $3/$15 becomes the cheaper rate card. Per completed task at the high setting, the two cost about the same today, and Neon costs about half after December 1.

The decision in three lines:

  • Switch now if Carbon's long reasoning runs are what slow you down. Neon High matches Carbon High on the index and finishes AA's index tasks in 40% of the time.
  • Switch by December 1 if cost per task is what you track. After the price change, Neon High's measured cost per task is about half of Carbon High's.
  • Stay on Carbon for your hardest reasoning. It still leads on research-level physics and has the lower hallucination rate, and Google keeps it for the cases where "a faster model gets stuck".

What Neon is, and where Google puts it

The model ID is gemini-4-neon, the same on AIHubMix. Google lists it as stable, released November 18, 2026. It has a March 2026 knowledge cutoff, a 1,048,576-token context window shared by input and output, and text, image, video, audio and PDF input.

Google places it between the fast and the deep model in the family. The AI Studio notes say Neon "sits between Gemini 4 Hydrogen and Gemini 4 Carbon for high-frequency complex tasks." The model-selection guide is more specific: Neon "finishes faster with fewer output tokens than Carbon while handling harder reasoning than Hydrogen," and Carbon remains "our strongest reasoning model". Google's speed labels for the two are "Moderate" for Neon and "Deliberate" for Carbon.

Neither model is on Google's free tier.

What changed from Carbon

Item Neon Carbon
Price, input / output $3 / $15 $2 / $10, then $4 / $20
Thinking levels low, medium, high, xhigh low, medium, high
Default level high medium
Output per request 65,536 65,536
Native non-streaming request 65,536 32,768
Output speed (AA) about 100 tok/s about 55 tok/s

Prices are Google's, in USD per million tokens; Carbon's $2/$10 runs through November 30, 2026, and $4/$20 applies from December 1. The non-streaming row is the output a single non-streaming REST call returns on Google's own API before it needs a continuation. Output speed is Artificial Analysis's measurement of Google's API at the high setting.

Two of these rows change behaviour without a code change, and one adds an option.

The default level is one step higher. Carbon defaults to medium; Neon defaults to high. A Carbon integration that never set a level gets more thinking per turn on Neon. On quality that's the comparable setting (see below), but it isn't the same token budget.

Non-streaming answers can run twice as long. A direct non-streaming call to Carbon returns at most 32,768 output tokens. Google's long-decode guide says that setting max_output_tokens above 32,768 on such a call "returns a 400 error". Neon allows 65,536 either way. AIHubMix's parameter schema records the same split: Carbon at 65,536 streaming and 32,768 non-streaming, Neon at 65,536 for both. A Carbon integration that kept non-streaming calls under 32K needs no change; one that moves to Neon can raise the cap.

There is a fourth level. xhigh is new for this tier; Carbon tops out at high. It only applies when a request asks for it. Google's advice is to "reserve xhigh for hard, multi-step problems where high falls short."

The rest is the same family design: one unified 1M window, implicit caching only, ignored sampling parameters and the same request schema. Google's checklist for moving from another Gemini 4 model has three lines: change the model ID, check the thinking level, use a paid project. The details that do break are in this series' migration post.

Quality: level at the defaults, different at the edges

The figures in this post come from Artificial Analysis (AA): its live leaderboard and model pages, and its coding-agent index, all re-checked 2026-11-19 at 13:27 UTC. They are third-party measurements of Google's API and change as AA re-runs models.

Setting AA Index Min/task First answer (s)
Neon Xhigh 61.4 21.5 56.5
Carbon High 60.0 36.6 72.9
Neon High (default) 59.6 14.8 40.6
Carbon Medium (default) 58.3 29.3 32.2
Neon Medium 54.9 8.3 20.1
Carbon Low 54.2 29.8 8.7
Neon Low 44.1 3.7 13.9

"First answer" is AA's time to first answer token. "Min/task" is AA's average time per Intelligence Index task.

At the same setting the scores are within half a point: 59.6 for Neon High, 60.0 for Carbon High. Neon's default scores 1.3 points above Carbon's default. Neon Xhigh is the highest Gemini 4 setting on the index, at 61.4.

Google's launch post calls that "#3 overall". It holds when each model counts once: only Claude Fable 5.2 and GPT-6.2 Astra score higher. AA ranks every setting separately, and there Neon Xhigh is sixth.

The average hides where the two differ. On AA's individual evaluations, Neon High beats Carbon High on FrontierCode (43.3 vs 40.6), Terminal-Bench-Science (65.7 vs 64.8), AutomationBench (79.2 vs 77.8) and the AA-Briefcase knowledge-work rating (1632 vs 1611). Carbon High keeps three clear leads that matter for hard work, plus a narrow one on HLE-Diamond (69.1 vs 68.4):

  • CritPt, AA's research-level physics evaluation: 69.0 against 59.0. No Neon setting scores above 62.
  • Terminal-Bench 4.0: 65.2 against 61.6.
  • Non-hallucination rate on AA's Omniscience test: 88.6% against 86.1%, and 95.8% for Carbon Medium.

AA's coding-agent leaderboard runs each model inside a coding agent; Google's models run in its Antigravity CLI. There, Neon Xhigh is first at 58.6, ahead of Carbon High at 55.5. Neon's lead over second place, Claude Code with Fable 5.2 Xhigh, is less than 0.2 points: a tie at the top, not a win.

Google's model card doesn't settle this comparison: its table sets Neon against Claude Opus 5.5 and GPT-6.2 Sol, not against Carbon.

Price and real cost

Model and period Input Output Cached AIHubMix today
Neon $3.00 $15.00 $0.15 same
Carbon, to Nov 30 $2.00 $10.00 $0.10 same
Carbon, from Dec 1 $4.00 $20.00 $0.20 not published

Google's Gemini API pricing, USD per million tokens, standard tier. Output includes thinking tokens. Neon has one rate across the full 1M window and no introductory period.

Per token, Neon costs 50% more than Carbon until November 30 and 25% less from December 1. AIHubMix lists both models at Google's current rates. It hasn't published a Carbon price for December, so check the Carbon model page then.

Per token is the wrong unit for these two, because they spend very different numbers of tokens on the same job. AA's cost per Intelligence Index task is a better measure.

Setting $/task to Nov 30 $/task from Dec 1
Neon Medium 1.90 1.90
Neon High (default) 3.74 3.74
Neon Xhigh 6.44 6.44
Carbon Medium (default) 3.29 6.59
Carbon High 3.88 7.76

AA's cost per task in USD, same snapshot. AA prices Carbon at its December rate ($4/$20). Every Carbon rate is exactly half that until November 30, so the two Carbon figures in the first column are AA's numbers halved: computed estimates, not AA measurements.

Read across the rows:

  • Today, Neon High and Carbon High cost about the same per task, $3.74 against about $3.88. Carbon Medium, at about $3.29, is cheaper than Neon High. Neon's advantage before December is time, not money.
  • From December 1, Neon High costs 52% less per task than Carbon High and 43% less than Carbon Medium, at a similar or higher score.

Google's launch post claims "a 32% lower cost per task than Carbon" at Neon's default level. Today's AA numbers give 43–52% after December 1, and anywhere from 3.5% cheaper to 14% dearer before it. Treat Google's figure as Google's own estimate.

A worked bill

Take one agent turn with a 200,000-token prompt. 150,000 of those tokens are a stable prefix that hits the implicit cache, and the model writes 20,000 tokens of thinking and answer.

  • Neon: (50,000 × $3 + 150,000 × $0.15 + 20,000 × $15) per million = $0.47.
  • Carbon until November 30: $0.32.
  • Carbon from December 1: $0.63.

That holds the token counts equal, which flatters Carbon against Neon High but not against Neon Xhigh. AA's per-task figures imply that at equal prices, Neon High would cost about 36% less than Carbon High on the same tasks, and Neon Xhigh about 11% more.

Two settings decide your real number: the thinking level you send, and whether your stable prefix clears the 4,096-token minimum for implicit caching. AA's model pages say Neon Xhigh generated 283 million tokens across the Intelligence Index, and Carbon High 205 million, against a 65.1 million median.

Speed: the reason to switch before December

On long work, Carbon is slow. An average Intelligence Index task takes 36.6 minutes at High and 29.3 at Medium. Neon High takes 14.8 minutes, and Neon Medium 8.3. Output speed explains most of it: AA measures Neon at about 100 tokens per second and Carbon at about 55.

The first answer token is a different clock. Carbon High makes you wait 72.9 seconds, and Neon High cuts that to 40.6. But Carbon's default Medium starts in 32.2 seconds, faster than Neon High. For chat-style use, compare Carbon Medium with Neon Medium: 20.1 seconds to the first answer and 8.3 minutes per task, at 3.4 points lower on the index.

On the coding-agent leaderboard the time gap is much smaller: Neon Xhigh averaged 30.5 minutes per task against 33.6 for Carbon High, about 9% less. Agent runs also spend time executing tools, which output speed doesn't change. AA puts those runs at $6.86 and $9.75 per task, pricing Carbon at its December rate; at today's rate Carbon High would be about $4.88, cheaper than Neon Xhigh.

Where Carbon is still the better choice

The hardest reasoning. CritPt is the clearest gap: 69.0 for Carbon against 59.0 for Neon High. Google's guidance points the same way: use Carbon for "system architecture, multi-file debugging, dense multimodal analysis, formal proofs, and security audits, or when a faster model gets stuck."

Fewer made-up answers. On AA's non-hallucination rate, Carbon Medium scores 95.8% against 86.1% for Neon High. If a confident wrong answer costs more than a slow right one, that gap matters more than the index.

This month's bill, for Medium users. Until November 30, Carbon Medium costs less per task than Neon High. If your evaluations pass on Carbon Medium and nobody is waiting on the answers, staying until the price change costs nothing.

A quick first token. Carbon Low returns its first answer token in 8.7 seconds, the fastest of any setting here, and scores 54.2. It still takes 29.8 minutes per task, so it helps only where the first token is the metric.

Who should switch now, who should wait

Switch now if your Carbon runs hit client timeouts, or if long reasoning tasks are the bottleneck. Start at Neon's default high: it matches Carbon on the index in half of Carbon Medium's time per index task. For tool-heavy agent loops the time saving is smaller, so measure your own runs. For interactive turns, use medium: it starts answering 12 seconds sooner than Carbon Medium, at 3.4 points lower on the index.

Switch before December 1 if cost per task drives the decision. Run your own evaluation set on Neon High and on Carbon in the last week of November, and move on the day the prices change.

Stay on Carbon for research-grade reasoning, security audits and work where the hallucination rate is the deciding number. You can also run both: Neon for the bulk of complex work, Carbon for the cases Neon gets wrong.

Set the level on purpose. A Carbon integration that never set a thinking level will run at high on Neon. That matches Carbon High, and costs twice as much per task as Neon Medium.

Try both through one endpoint

Both models are on AIHubMix's OpenAI-compatible endpoint, so switching is a one-line change. AIHubMix's parameter schema lists reasoning_effort values from low to xhigh for Neon, and low to high for Carbon.

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://aihubmix.com/v1",
    api_key=os.environ["AIHUBMIX_API_KEY"],
)

for model in ["gemini-4-neon", "gemini-4-carbon"]:
    response = client.chat.completions.create(
        model=model,
        reasoning_effort="high",  # Neon also accepts "xhigh"; Carbon tops out at "high"
        max_completion_tokens=32768,  # Carbon's non-streaming limit; Neon allows 65,536
        timeout=900,  # hard prompts can take several minutes
        messages=[{"role": "user", "content": "Find the race condition in this retry queue and propose a fix: ..."}],
    )
    print(model, response.usage.completion_tokens)
    print(response.choices[0].message.content[:500])

The sample stays under 32,768 output tokens because it doesn't stream, and that's Carbon's non-streaming limit; stream the response to go up to 65,536 on Carbon. AIHubMix's OpenAI-compatible endpoint can't continue past 65,536. Its schema lists Google's continuation token on the native Gemini endpoint, which this post didn't test. No live test was run for this post, so the sample shows the request shape, not a verified response. The compare page puts the two model listings side by side.

FAQ

Is Gemini 4 Neon better than Carbon?
At the same setting they score about the same on the Artificial Analysis Intelligence Index: 59.6 for Neon High and 60.0 for Carbon High. Neon finishes tasks in well under half the time. Carbon still leads on research-level reasoning such as CritPt and has a lower hallucination rate.

Is Neon cheaper than Carbon?
Not per token until November 30: Neon is $3/$15 and Carbon $2/$10. From December 1 Carbon is $4/$20, so Neon is 25% cheaper per token. Per task at the high setting, AA's figures put Neon at about half of Carbon from December 1, and about level with Carbon today once Carbon's cost is halved to its current rate.

What is Neon's default thinking level?
High. Carbon's default is medium. Neon also adds xhigh, which Carbon doesn't have.

Do I need to change my requests to move from Carbon to Neon?
Google says no request schema changes are needed. Change the model ID, and set the thinking level explicitly if you relied on Carbon's medium default.

Is Neon faster in every case?
No. Per task it is: 14.8 minutes for Neon High against 36.6 for Carbon High on AA. To the first answer token, Neon High (40.6 seconds) is slower than Carbon's default Medium (32.2). Neon Medium (20.1 seconds) is faster than both.

What will AIHubMix charge for Carbon after December 1?
It hasn't been published. AIHubMix lists Carbon at Google's current $2/$10 today, and Google's rate goes to $4/$20 on December 1. Check the Carbon model page then.

Can Neon generate more than 65,536 tokens in one response?
On Google's own API, yes, through long decode in the Interactions API or GenerateContent. Through AIHubMix's OpenAI-compatible endpoint the limit is 65,536 output tokens per request, and 32,768 for non-streaming Carbon calls.

Keep reading: the Gemini 4 Neon series

Sources

One gateway to multiple AI providers. Built for production.

Visit AIHubMix