cd /news/large-language-models/googles-faster-gemini-4-neon-edges-p… · home › topics › large-language-models › article
[ARTICLE · art-172412] src=madrobot.blog ↗ pub= topic=large-language-models verified=true sentiment=· neutral

Google’s faster Gemini 4 Neon edges past its own flagship on a top AI leaderboard

Google released Gemini 4 Neon on Wednesday, a faster model adapted from flagship Gemini 4 Carbon priced at $3 per million input tokens and $15 per million output tokens, scoring 61 on Artificial Analysis's independent index — one point above Carbon — while completing tasks in about 40% less time. Neon is the fourth Gemini 4 model after Argon, Carbon and Hydrogen, reads text, images, audio and video, carries a 1 million token context window and a March 2026 knowledge cutoff, and is available now to Google AI Pro and Ultra subscribers in the Gemini app. Artificial Analysis called Neon "very verbose," using 283 million tokens on its tests against a median of 65.1 million, and "notably slow per task" next to similarly priced models, while Google says a task at Neon's default setting costs 32% less than on Carbon.

key takeaway

Gemini 4 Neon scores 61 on Artificial Analysis's index versus Carbon's 60 at $3/$15 per million tokens, but burns 283 million tokens on the test suite against a 65.1 million median, so budget for verbosity rather than assuming the lower per-token price cuts total task cost.

by read4 min views10 publishedNov 19, 2026
Google’s faster Gemini 4 Neon edges past its own flagship on a top AI leaderboard
Image: Madrobot (auto-discovered)

Google’s launch artwork for Gemini 4 Neon. Image: Google

Google released Gemini 4 Neon on Wednesday, a faster model built from its flagship Gemini 4 Carbon, at $3 per million input tokens and $15 per million output tokens. On Artificial Analysis’s independent index, Neon at its top setting scores 61, one point above Carbon, and finishes its tasks in about 40% less time. Google AI Pro and Ultra subscribers can pick it in the Gemini app now.

What Gemini 4 Neon is #

Neon is the fourth Gemini 4 model, after Argon in September, Carbon and, last week, the cheap Hydrogen. Its model card says it is “adapted from Gemini 4 Carbon”. It reads text, images, audio and video, has a 1 million token context window and a March 2026 knowledge cutoff.

Google pitches it as the middle of the range: “While Carbon tackles your hardest, long-horizon projects and Hydrogen handles fast everyday execution, Neon balances depth and efficiency.” It has four thinking levels, with high as the default. In Google’s Antigravity coding tool it can also act as an orchestrator, directing several Hydrogen agents working in parallel.

How Neon compares with Hydrogen, Carbon and Argon #

Artificial Analysis runs every model through the same set of tests and publishes the score, the cost and the speed:

Model (setting) API price per 1M tokens (in/out) Index score Cost per task Output speed
Gemini 4 Neon (xhigh) $3 / $15 61 $6.44 110 tokens/s
Gemini 4 Carbon (high) $2 / $10, then $4 / $20 from Dec 1 60 $7.76 55 tokens/s
Gemini 4 Hydrogen (xhigh) $1.50 / $7.50 54 $3.16 169 tokens/s
Gemini 4 Argon (high) $2 / $10, then $4 / $20 from Dec 1 50 $4.97 65 tokens/s

Artificial Analysis Intelligence Index and API prices as listed on November 19, 2026. Its cost per task uses Carbon’s December price and Argon’s current introductory price.

So Neon is twice as fast as Carbon and, once Carbon’s introductory pricing ends on November 30, a quarter cheaper per token. Until then Carbon is the cheaper model per token. Google says that at Neon’s default setting a task costs 32% less than on Carbon. Artificial Analysis is less flattering on efficiency: it calls Neon “very verbose”, using 283 million tokens on its tests against a median of 65.1 million, and “notably slow per task” next to models at a similar price.

Unlike Hydrogen, Neon has no free API tier, according to Google’s pricing page. Batch requests cost half, $1.50 and $7.50, and cached input is 95% cheaper.

Google’s benchmark claims #

Google says Neon ranks third overall on Artificial Analysis’s index, behind what it calls “larger, slower models”, a group that includes Claude Fable 5.2. Its model card compares Neon with Claude Opus 5.5 and GPT-6.2 Sol, not Fable. These are Google’s own figures:

Test Gemini 4 Neon Claude Opus 5.5 GPT-6.2 Sol
Vals Index v2.2 (professional tasks) 70.9% 70.1% 66.1%
Zapier AutomationBench 56.3% 42.5% 41.3%
Terminal-Bench 4.1 (coding) 65.2% 65.8% 61.5%
FrontierCode 2.0 (coding) 46.2% 47.3% 43.8%
Chartography (reading charts) 73.5% 66.3% 75.1%

On safety, Google says alerts from its monitor for misaligned reasoning and actions stayed below Carbon’s already low rates, and that Neon took shortcuts on 2.45% of its coding cheating tests, the lowest of any Gemini 4 model. Its full safety report is due “in the coming days”. Outside checks have been harsher: Redwood Research said this week that Argon slipped past a realistic monitor 45.5% of the time.

Who gets Gemini 4 Neon, and where #

  • Gemini app: Google AI Pro ($19.99 a month in the US) and Ultra subscribers can pick Neon from the model menu, according to Google’splans page . Free users don’t get it.
  • Developers: the paid Gemini API tier in Google AI Studio and Android Studio, plus Antigravity for paying Pro and Ultra subscribers. It is also listed onOpenRouter at the same price.
  • Businesses: Gemini for business, through the API or its model menu, and Antigravity.
  • Security teams: a “Neon for cyber” version reaches members of Google’s Fairwind Program over the next few days.

Google AI Pro is sold in more than 150 countries and territories, including the US, UK, Canada, Germany, France, India and Australia, according to Google’s list. Google hasn’t named any country where Neon is missing.

Why it matters #

Neon gives paying users near-flagship results at roughly twice the speed, and on one independent index it edges past Carbon itself. Four Gemini 4 models in seven weeks show how fast Google is now shipping, though Artificial Analysis’s numbers suggest its speed comes with a heavy token bill.

Sources: Google (announcement), Gemini 4 Neon model card, Gemini API pricing, Google AI plans, Artificial Analysis, OpenRouter.

── more in #large-language-models 4 stories · sorted by recency
── more on @google 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
→ Live at https://your-agent.zahid.host ✓
Get free account → Pricing
from €0/mo · no card required
LIVE [news/googles-faster-gemin…] indexed:0 read:4min 2026-11-19 · —