Google
Released November 18, 2026Cutoff March 2026

Gemini 4 Neon

Gemini 4 Neon by Google pairs frontier reasoning with fast end-to-end task completion, a 1M-token context, Hydrogen subagent orchestration, and cyber defense.

Inputs
Text
Image
File
Audio
Video
Outputs
Text

Model Overview

Capabilities, design details, and architectural traits

Gemini 4 Neon - Frontier intelligence that balances depth and efficiency

Gemini 4 Neon is Google's model for complex, high-frequency workflows across software engineering, knowledge work, and science. It is adapted from Gemini 4 Carbon and sits deliberately between the Gemini 4 lineup: Carbon takes the hardest long-horizon projects, Hydrogen handles fast everyday execution, and Neon delivers frontier-level reasoning faster.

DifferentiatorWhat it does
Balanced lineup roleAdapted from Gemini 4 Carbon; early evaluations matched or exceeded Carbon on overall quality at 3x the speed.
Orchestrator modeRuns standalone or acts as a fast, high-judgment Orchestrator directing parallel Hydrogen subagents in Google Antigravity, including /swarm agent swarms.
1M unified contextA 1 million-token context window spanning input and output, with long decode up to 1M tokens.
Four thinking levelsLevels scale steadily from the default high setting to xhigh, which adds deeper multi-step reasoning for the hardest engineering, analytical, and scientific workloads.
Cyber defense lifecycleCovers secure code generation, penetration testing, and vulnerability finding, validating, and patching.
Multimodal knowledge workSynthesizes spreadsheets, PDFs, charts, audio, and video in a single context window and turns them into structured deliverables.
Long-session disciplineCuts redundant file reads, shares progress as it works, and handles context compaction reliably in long agent sessions.

Built for production-ready code

Neon is engineered for long-running, tool-heavy agent workflows and sustains performance end-to-end. It handles multi-file debugging and refactoring across repositories with targeted, surgical edits, and external testers highlighted its ability to explore large code bases and find subtle vulnerabilities other models missed.

Efficiency and control

The thinking levels let teams dial intelligence per workload. At its default high level, Neon sits on the Pareto frontier on both the Vals Index and Agent Arena.

Benchmark Performance

Independent evaluations · Artificial Analysis

61.4%
Intelligence

Compare Models Side-by-Side

Evaluate specifications, pricing, and independent benchmark indices

Model Details
General Info
ProviderGoogleAnthropicOpenAI
Release DateNovember 18, 2026November 9, 2026November 17, 2026
Knowledge CutoffMar 2026--
Context & Limits
Context Window1.0M1M
1.1M
Best Context Window
Pricing (per 1M tokens)
Input Pricing
$3
Best Input Pricing
$8$8
Output Pricing
$15
Best Output Pricing
$40$40
Modalities
Inputs
textimagefileaudiovideo
textimagefile
textimagefile
Outputs
text
text
text
Benchmarks (0-100)
Intelligence Index61.4
64.6
Best Intelligence Index
62.1
Coding Index---
Agentic Index---

Benchmark comparison not available

Select at least two models that have available benchmark evaluation data to render the performance comparison charts.

Independent evaluation data provided by Artificial Analysis. To view the latest benchmarks and full details, visit their official site.

Explore more from Google

Other models by Google

Top AI Models

Leading alternatives by intelligence score

View all