Subscribe
0%
Illustration for: Google's Gemini 4 Neon: a $3/$15 mid-tier model that tops Opus 5.5 on Google's charts. What it costs and what's unverified

Google's Gemini 4 Neon: a $3/$15 mid-tier model that tops Opus 5.5 on Google's charts. What it costs and what's unverified

Google released Gemini 4 Neon on November 18, the fourth Gemini 4 model in about a month. It sits between Hydrogen and Carbon, costs $3/$15 per million tokens and ships in the API and the Gemini app. Google's own tables put it ahead of Claude Opus 5.5 and GPT-6.2 Sol on most rows. The cost-per-task claims don't line up across reports.

Listen to this article0:00 / 5:05

Google released Gemini 4 Neon on November 18. It's a mid-tier model that sits between the fast, cheap Hydrogen and the flagship Carbon. According to The Decoder, it's the fourth new Gemini model in about a month. Google pitches it as "frontier-level intelligence" for software engineering, knowledge work and science. It went live the same day in the paid Gemini API and for Google AI Pro and Ultra subscribers in the Gemini app.

Mid-tier models are where most production agent traffic ends up, so this matters for cost. Google's model-card tables, as relayed by several outlets, show Neon ahead of Anthropic's Claude Opus 5.5 and OpenAI's GPT-6.2 Sol on most rows, at a list price between the two. Nearly every number below comes from Google, and the cost-per-task claims differ from report to report.

Price and specs

Gemini 4 Neon Source
Input / output (per 1M tokens) $3 / $15 SiliconANGLE, Decoder, Digital Trends
Cached input $0.15 SiliconANGLE, BigGo
Free API tier None (Hydrogen keeps one) BigGo
Context ~1M tokens Decoder, Digital Trends
Output 64K synchronous; up to 1M via background long-decode mode BigGo (Decoder: "up to 1M")
Thinking levels Four, "high" by default Digital Trends, BigGo
Knowledge cutoff March 2026 BigGo
Inputs Text, images, video, audio, PDFs (no image/audio generation) Decoder
Tools Computer Use, code execution, file search, Search and Maps grounding Decoder

For comparison, SiliconANGLE lists GPT-6.2 Sol at $2/$10 and Claude Opus 5.5 at $4/$20. Carbon's pricing is the catch. Carbon is on an introductory $2/$10 until November 30 and moves to $4/$20 on December 1, per The Decoder and BigGo. That means Neon costs more per token than Carbon for the next two weeks and less after that.

Where you can use it

  • Gemini app: AI Pro and Ultra subscribers get it under "Deeper analysis," where it replaces Argon in the model picker (Digital Trends). BigGo notes that AI Pro is $19.99/month in the US, while Carbon in the app needs Ultra, which starts at $99.99.
  • Developers: the paid tier of the Gemini API in AI Studio, plus Android Studio and Google Antigravity for Pro/Ultra subscribers. In Antigravity, an Orchestrator mode has Neon plan while Hydrogen executes.
  • Enterprise: Gemini for business. Digital Trends says the Vertex AI documentation lists Neon for verified enterprise customers in global, US and EU regions.
  • Cyber variant: BigGo reports that a security-tuned version is going to vetted defenders through Google's Fairwind Program "in the coming days." Carbon used the same defenders-first route.

Google's benchmark table

These are Google's model-card figures as reported by BigGo, with matching rows confirmed by Digital Trends. We have not seen the underlying runs.

Benchmark Neon Opus 5.5 GPT-6.2 Sol
Artificial Analysis Intelligence Index 61 59 57
Vals Index 70.9% 70.1% 66.1%
Terminal-Bench 4.1 65.2% 65.8% 61.5%
FrontierCode 2.0 46.2% 47.3% 43.8%
AutomationBench 56.3% 42.5% 41.3%
Chartography 73.5% 66.3% 75.1%
GraphWalks (256K–1M) 95.2% 66.8% 80.1%

Neon loses three rows (Terminal-Bench, FrontierCode and Chartography) by 0.6 to 1.6 points. Its biggest leads come on long-context graph traversal and automation tasks. BigGo also reports a "SWE-sweep" score of 39.0% against Opus's 10.1%. That run used a provider-specific harness, so we left it out of the table.

SiliconANGLE reports that Artificial Analysis ranks Neon third overall on its Intelligence Index at 61, behind Claude Fable 5.2 (65) and GPT-6.2 Astra (62). So Neon is a strong mid-tier model, not a new overall leader.

The cost claims don't agree

  • Google, per Digital Trends and BigGo: at default "high" thinking, Neon roughly matches Carbon on the index at 32% lower cost per task, and runs 50–80% cheaper per task than leading frontier models.
  • SiliconANGLE: says Neon finishes most tasks about 3x faster than Carbon and costs "around 60% less." It also cites Artificial Analysis putting Neon's cost per task only about 17% below Carbon's.

These figures can't all describe the same comparison. Some are probably per-token and some per-task, and they may use different thinking levels or Carbon's intro versus full price. None of the reports we read reconcile them. Until Artificial Analysis publishes its full cost breakdown, treat 17% as the conservative independent number and 32% or more as Google's claim.

What to do

  • If you're on Carbon for agent workloads, test Neon before December 1, when Carbon's price doubles. If Neon holds up on your evals, it's the cheaper per-token option from that date.
  • If you're on Opus 5.5 or GPT-6.2 Sol, Google's tables show Neon close or ahead on most tasks. Coding-agent benchmarks are the exception, where Opus still edges it. Run your own harness, because vendor numbers come from vendor setups.
  • Free-tier API users stay on Hydrogen, since Neon has no free tier.
  • Watch for Google's full model card, independent Vals and Artificial Analysis cost data, and the cyber variant's rollout.

Google's own launch post was not readable for this article, so the Google claims above come via SiliconANGLE, The Decoder, Digital Trends and BigGo.

Sources

  1. SiliconANGLE — Google's latest midtier model Gemini 4 Neon is three times faster than Carbon and wins on some metrics siliconangle.com
  2. The Decoder — Google ships Gemini 4 Neon, its fourth new Gemini model in a month the-decoder.com
  3. Digital Trends — Google says Gemini 4 Neon can hang with Claude Opus 5.5 and GPT-6.2 Sol for less digitaltrends.com
  4. BigGo Finance — Google Unveils Gemini 4 Neon, a Cheaper Frontier Model Aimed at Coding and Enterprise Workloads finance.biggo.com

#Google#Gemini#Gemini 4 Neon#Gemini 4 Carbon#Gemini 4 Hydrogen#Antigravity#Gemini API#Artificial Analysis#pricing#benchmarks