Summary:

Google’s new Gemini model costs $3 per million input tokens. Compare its rollout and benchmarks with Claude Opus 5.5 and GPT-6.2 Sol.

Google began rolling out Gemini 4 Neon on November 18. The model sits between Hydrogen, which Google positions for fast everyday tasks, and Carbon, which it recommends for longer and harder projects. Neon is aimed at frequent work that still calls for deeper reasoning, such as revising a codebase or building an analysis from documents, charts and spreadsheets.

Price and access

Google lists Neon at $3 per million input tokens, $0.15 per million cached input tokens and $15 per million output tokens. The context window is 1 million tokens. At those published rates, 100,000 uncached input tokens and 10,000 output tokens would cost $0.45 in token charges.

The developer rollout starts in the paid Gemini API tier via Google AI Studio and Android Studio. Gemini for business is getting Neon in the API and model picker. In the Gemini app and Google Antigravity, access is rolling out to paid Google AI Pro and Ultra subscribers. The rollout does not mean the model is already selectable on every eligible account.

How Neon compares with Claude Opus 5.5 and GPT-6.2 Sol

These selected rows use the scores in Google DeepMind’s November 2026 comparison table. Artificial Analysis and Vals publish their own benchmark indices; the comparison as presented here is sourced to DeepMind. Google cites its xhigh thinking setting for Neon’s score of 61 on the Artificial Analysis index. Higher is better in these rows; the highest score in each row is bolded.

BenchmarkGemini 4 NeonClaude Opus 5.5GPT-6.2 Sol
Artificial Analysis Index v5.0.1615957
Vals Index v2.270.9%70.1%66.1%
Terminal-Bench 4.1 (adjusted accuracy)65.2%65.8%61.5%
FrontierCode 2.0 (main score)46.2%47.3%43.8%
AutomationBench56.3%42.5%41.3%
Chartography73.5%66.3%75.1%

Neon leads these three models on both aggregate indices and AutomationBench. Claude Opus 5.5 is slightly ahead on the two selected coding rows, and GPT-6.2 Sol leads Chartography. The benchmarks use different tasks and scoring methods, so a higher aggregate score does not automatically make Neon the better fit for a terminal or chart-heavy workflow.

What to check before changing models

Pick a few repeatable jobs where your current model struggles, then compare successful completions, the number of corrections you make, time to completion and total token cost. Keep the input and scoring rules fixed. If a task normally uses a cached document, measure that separately from a first run: the published cached and uncached input prices differ by a factor of 20.

For a faster option within Google’s lineup, compare the Gemini 4 Hydrogen rollout and benchmarks.

Sources