> ## Content Index
> Fetch the complete content index at: https://claudefox.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# Gemini 4 Neon vs Claude Opus 5.5 and Sonnet 5.5: price, per-task cost and when to switch
- URL: https://claudefox.com/gemini-4-neon-vs-claude-opus-5-5-sonnet-5-5/
- Published: 2026-11-19T03:44:11.000Z
- Updated: 2026-11-19T03:44:23.000Z
- Description: Google released Gemini 4 Neon on November 18 at $3 input and $15 output per million tokens, priced between Claude Sonnet 5.5 and Opus 5.5. On Artificial Analysis it scores 61 at xhigh for $6.44 per task. Here is how it compares, and why the answer changes on December 1.
- Author: hupicake
- Tags: Research

Google released Gemini 4 Neon (model code gemini-4-neon) on November 18, 2026\. It sits between Gemini 4 Hydrogen and Gemini 4 Carbon, costs $3 per million input tokens and $15 per million output tokens, and reads up to 1 million tokens. Google says it scores 61 on the Artificial Analysis Intelligence Index at its xhigh thinking level, third overall (Artificial Analysis's own page listed it 6th of 96 on November 19). On the Claude side, its price sits between Claude Sonnet 5.5 ($2/$10) and Claude Opus 5.5 ($4/$20).

We have not tested Neon ourselves. This page compares published figures: Google's announcement and price page, Anthropic's price page, and Artificial Analysis, an independent benchmarking site. Neon's Artificial Analysis numbers were read on November 19; the Claude and Carbon numbers on November 18.

## Quick facts

- Model code: gemini-4-neon (stable). Inputs: text, image, video, audio and PDF. Output: text.
- Context window: 1,048,576 tokens, shared by input and output. Default synchronous output cap: 65,536 tokens; up to about 1 million with Google's long-decode mode.
- Price (Gemini API paid tier): $3 input, $15 output per million tokens, thinking tokens billed as output. Cached input $0.15 per million, implicit caching only. No free tier.
- Thinking levels: low, medium, high and xhigh. Default: high.
- Also available: Batch, Flex and Priority inference, Google AI Studio, Android Studio, Google Antigravity for paid Google AI Pro and Ultra subscribers, the Gemini app for Pro and Ultra subscribers, and Gemini for business.

## How much does Gemini 4 Neon cost compared with Opus 5.5 and Sonnet 5.5?

Per token, Neon costs 25% less than Opus 5.5 and 50% more than Sonnet 5.5\. List prices per million tokens:

- Claude Sonnet 5.5: $2 input, $10 output.
- Gemini 4 Neon: $3 input, $15 output.
- Claude Opus 5.5: $4 input, $20 output.
- Gemini 4 Carbon and Gemini 4 Argon: $2 input, $10 output through November 30, 2026; $4 and $20 from December 1.

Our arithmetic for one job that reads 200,000 tokens and writes 20,000, with no caching: Sonnet 5.5 $0.60, Neon $0.90, Opus 5.5 $1.20\. Carbon costs $0.60 until November 30 and $1.20 from December 1\. So through November 30 Google's top model is cheaper per token than Neon; after that, Neon is the cheaper of the two by 25%, a point 9to5Google also made.

## What does each model cost per task in independent tests?

Per-token price is only half of it, because models spend very different numbers of tokens on the same work. Artificial Analysis runs every model through the same set of tasks and reports the score, the cost and the time per task:

- Gemini 4 Neon (xhigh): score 61; $6.44 per task; 21.5 minutes per task; 110 output tokens per second; 56.5 seconds to the first answer token; 283 million output tokens across the index.
- Gemini 4 Carbon (high): score 60; $5.48 per task at the $2/$10 launch price; 33.2 minutes per task; 62 output tokens per second.
- Claude Opus 5.5 (Max setting): score 59; $10.04 per task; 24.5 minutes per task; 118 output tokens per second.
- Claude Sonnet 5.5 (Max setting): score 55; $10.02 per task; 32.3 minutes per task; 147 output tokens per second.

On these numbers Neon at xhigh scores two points above Opus 5.5 at max for about 64% of the cost per task, and finishes tasks about three minutes faster. Sonnet 5.5 is cheaper per token than Neon but cost more per task in this test, because it read and wrote more tokens. Artificial Analysis also flags Neon as very verbose: 283 million output tokens against a median of 65.1 million for comparable models.

Two cautions. These are one lab's tasks, not your codebase. And the Carbon figure uses its launch price; at December prices the same Carbon runs would come to about $10.96 per task, so Neon at xhigh ($6.44) costs more per task than Carbon today but less from December 1\. Google separately says Neon at its default high level costs 32% less per task than Carbon, without saying which Carbon price it used.

## What Google claims for coding

- Artificial Analysis Coding Agent Index: 58.6, first overall, per Google.
- Terminal-Bench 4.1: 65.2%.
- FrontierCode 2.0: 46.2% at the default high setting.
- SWE-sweep: 12.8% on the mini-SWE harness and 39.0% inside Google Antigravity.
- Google says Neon cuts redundant file reads, reports progress while it works, handles context compaction in long sessions, and can act as an orchestrator directing parallel Hydrogen subagents in Antigravity.

These are Google's chosen benchmarks and settings. We have not lined them up against Anthropic's tables here, because the two vendors often run different versions of the same benchmark (Terminal-Bench 4.0 versus 4.1, for example) and different effort settings. The Artificial Analysis index above is the like-for-like comparison.

One outside view: micjm, which runs the Agent Duel leaderboard, projected Neon's rating from Carbon's results plus Neon's speed and estimated 1,066 against Opus 5.5's 1,078, with a wide interval. That is a model of a leaderboard, not a measured result.

## How to try Gemini 4 Neon from the API

You need a paid-tier Gemini API key; Neon is not on the free tier. Set the model to gemini-4-neon and pick a thinking level with thinking\_level. Google's docs recommend starting at high and keeping xhigh for hard, multi-step problems where high falls short, because xhigh adds latency and tokens and can exhaust the output cap on simple prompts. This follows the Python example in Google's thinking docs (the Interactions API in the google-genai SDK), with the model name swapped:

```
from google import genai

client = genai.Client()

interaction = client.interactions.create(
    model="gemini-4-neon",
    input="Find the root cause of the failing test in this log: [paste log]",
    generation_config={
        "thinking_level": "high"
    }
)

print(interaction.output_text)
```

Every token is billed, including thinking tokens and the internal tool instructions Google adds when you use grounding. Grounding with Google Search gives 5,000 free requests a month shared across Gemini 3 and newer models, then $14 per 1,000, capped at 1,500 requests a day.

**Try it with your AI.** Before switching a workload from Claude to Gemini 4 Neon, have your AI build a small side-by-side cost and quality check on your own tasks.

```
I want to compare Claude [Opus 5.5 or Sonnet 5.5] with Gemini 4 Neon (model code
gemini-4-neon) on my own workload before switching.

First, check: do I have a paid-tier Gemini API key (Neon has no free tier) and a Claude
API key, both set as environment variables? Do not print or log either key.

Then write a script that:
1. Reads [5-10] sample tasks from [path to a folder of prompts and expected outputs].
2. Sends each task to both models. For Gemini, set thinking_level to high; for Claude,
   use [effort setting].
3. Sets a high max_output_tokens for Gemini and records any truncated or incomplete
   responses (thinking tokens are still billed).
4. Records input tokens, output tokens (including thinking tokens), wall-clock time and
   the raw answer.
5. Computes cost per task, assuming no caching, with these list prices per million
   tokens: Gemini 4 Neon $3 input / $15 output; Claude Opus 5.5 $4 / $20; Claude Sonnet
   5.5 $2 / $10.
6. Writes a table to results.md.

Show me the script and the estimated total cost before running it, and wait for my OK.
Do not install anything beyond the official Google and Anthropic SDKs, and do not delete
files.
```

## Where Neon sits in the Gemini 4 family

Google describes Carbon as the model for the hardest long-horizon projects, Hydrogen for fast everyday work, and Neon as the balance between them for complex coding and enterprise workflows. In the Gemini app, Neon replaces Argon in the model picker, according to 9to5Google. A cyber-tuned version, Gemini 4 Neon for cyber, is going to members of Google's Fairwind Program for security defenders over the coming days.

On safety, Google says Neon keeps Carbon-level safeguards against cyber and biosecurity misuse while refusing 1.98% of Terminal-Bench 4.0 tasks, and reports a 0.0% attack success rate on Gray Swan's indirect prompt injection benchmark at its highest thinking level with real-time safeguards on. Those are Google's own evaluations.

## Should you switch from Opus 5.5 or Sonnet 5.5?

- If cost per task matters and your work resembles what the Artificial Analysis index measures, Neon is worth an eval: higher index score than Opus 5.5 at about two-thirds of its per-task cost.
- If you mainly use Sonnet 5.5 for its per-token price, Neon is 50% dearer per token; whether it ends up cheaper depends on how many tokens each model spends on your tasks.
- If you are choosing between Neon and Carbon this month, Carbon is cheaper per token until November 30\. From December 1 Neon is 25% cheaper per token.
- If you depend on Claude-specific tooling such as Claude Code, keep in mind that switching models also means switching harness; Google's best SWE-sweep figure for Neon (39.0%) was measured inside its own Antigravity harness.

## FAQ

How much does Gemini 4 Neon cost? $3 per million input tokens and $15 per million output tokens on the Gemini API paid tier, with cached input at $0.15 per million. There is no free tier.

What is the Gemini 4 Neon model ID? gemini-4-neon. It has a 1,048,576-token context window and a default output cap of 65,536 tokens.

What thinking levels does Gemini 4 Neon support? Low, medium, high and xhigh, with high as the default. Google recommends starting at high and keeping xhigh for hard multi-step problems, because it adds latency and tokens.

Is Gemini 4 Neon better than Claude Opus 5.5? On the Artificial Analysis Intelligence Index, Neon at xhigh scored 61 and Opus 5.5 at max effort scored 59, with Neon at $6.44 per task against $10.04\. That is one independent benchmark; results on your own tasks can differ.

Is Gemini 4 Neon cheaper than Gemini 4 Carbon? Not per token until November 30, when Carbon is $2/$10\. From December 1 Carbon goes to $4/$20, and Neon at $3/$15 is 25% cheaper.

Can I use Gemini 4 Neon in the Gemini app? Yes, for Google AI Pro and Ultra subscribers, from the model drop-down. Paid Pro and Ultra subscribers can also use it in Google Antigravity.

## What we have not tested

We have not run Gemini 4 Neon on our own tasks, so nothing here is a hands-on result. Benchmark figures come from Google or Artificial Analysis and are dated above. If we add our own measurements, we will date them on this page.

## Related on claudefox

- [Gemini 4 Carbon vs Claude Opus 5.5 and Fable 5.2: benchmarks and pricing compared](https://claudefox.com/gemini-4-carbon-vs-claude-opus-5-5-fable-5-2/)
- [Gemini 4 Hydrogen vs Claude Sonnet 5.5 for coding: price, speed and benchmarks](https://claudefox.com/gemini-4-hydrogen-vs-claude-sonnet-5-5/)
- [GPT-6.2 benchmarks vs Claude Fable 5.2: what independent evaluators found](https://claudefox.com/gpt-6-2-benchmarks-vs-claude-fable-5-2/)
- [How to call Gemini 4 Hydrogen from Claude Code: gemini-api CLI, Docs MCP and skills](https://claudefox.com/gemini-4-hydrogen-claude-code-setup/)

Source: [Introducing Gemini 4 Neon (Google, Nov 18, 2026)](https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-neon/?ref=claudefox.com), [Gemini API pricing (Google)](https://ai.google.dev/gemini-api/docs/pricing?ref=claudefox.com), [Gemini 4 Neon model page (Google AI Studio docs)](https://ai.google.dev/gemini-api/docs/models/gemini-4-neon?ref=claudefox.com), [Gemini models overview (Google AI Studio docs)](https://ai.google.dev/gemini-api/docs/models?ref=claudefox.com), [Gemini thinking levels (Google AI Studio docs)](https://ai.google.dev/gemini-api/docs/thinking?ref=claudefox.com), [Gemini 4 Neon (Artificial Analysis)](https://artificialanalysis.ai/models/gemini-4-neon?ref=claudefox.com), [Claude Opus 5.5 (Artificial Analysis)](https://artificialanalysis.ai/models/claude-opus-5-5?ref=claudefox.com), [Claude Sonnet 5.5 (Artificial Analysis)](https://artificialanalysis.ai/models/claude-sonnet-5-5?ref=claudefox.com), [Gemini 4 Carbon (Artificial Analysis)](https://artificialanalysis.ai/models/gemini-4-carbon?ref=claudefox.com), [Claude API pricing (Anthropic docs)](https://platform.claude.com/docs/en/about-claude/pricing?ref=claudefox.com), [Gemini 4 Neon announcement coverage (9to5Google)](https://9to5google.com/2026/11/18/gemini-4-neon-announcement/?ref=claudefox.com), [Gemini 4 Neon looks like an Opus 5.5 competitor (micjm)](https://micjm.com/blog/gemini-4-neon-looks-like-a-top-model/?ref=claudefox.com)

This story first appeared in the [claudefox daily briefing for November 19, 2026](https://claudefox.com/daily-2026-11-19/).