---
title: "Google DeepMind launches Gemini 4 Neon model for enterprise and scientific workloads"
slug: "google-deepmind-launches-gemini-4-neon-model-for-enterprise-and-scient"
date: "2026-11-19"
category: "AI & Automation"
author: "Dr. Kai Wei"
correspondent: "Dr. Kai Wei"
correspondentRole: "Lead Frontier AI Correspondent"
readingTimeMinutes: 2
tags: ["AI","News"]
originalSourceUrl: "https://deepmind.google/blog/introducing-gemini-4-neon/"
canonicalUrl: "https://aurav.ae/google-deepmind-launches-gemini-4-neon-model-for-enterprise-and-scient"
publisher: "Aura-V Intelligence (aurav.ae)"
---

# Google DeepMind launches Gemini 4 Neon model for enterprise and scientific workloads

> Gemini 4 Neon offers a 1 million-token context window, lower pricing and top benchmark rankings across finance, legal, coding and science.

## Key Takeaways (TL;DR)

- Gemini 4 Neon is available today via the Gemini API, Google AI Studio, and paid Google AI Pro/Ultra subscriptions.
- The model costs $3 per million input tokens (95% discount on cached inputs) and $15 per million output tokens, with a 1 M-token context window.
- Independent tests place Neon at #2 on the Vals Index, #1 on Zapier’s AutomationBench and top of several scientific and cybersecurity benchmarks.

Google DeepMind announced Gemini 4 Neon, a new model that “brings frontier-level intelligence with fast end-to-end task completion” across software engineering, knowledge work and scientific workflows. The model is live today in the Gemini API, Google AI Studio, Gemini for business (API and model dropdown), and for paid Google AI Pro and Ultra users in Google Antigravity and the Gemini app.

Neon is priced at $3 per million input tokens (or $0.15 per million cached input tokens, a 95 % discount) and $15 per million output tokens, and supports a 1 million-token context window. Benchmark results show Neon ranking #2 on the Vals Index (70.9 %), #1 on Zapier’s AutomationBench (56.3 %) and top positions on legal, tax, finance and spreadsheet tasks, as well as best-in-class scores on Terminal-Bench Science (70.5 %), RiemannBench (84.0 %) and HLE-Diamond (71.3 %). In cybersecurity, Neon achieved state-of-the-art results on AutoBaxBench (hard tier 76 %) and WebExploitBench v1.2 (51.26 %) at roughly half the cost of competing large models.

The model is engineered for long-running agent workflows, offering reduced redundant file reads, context compaction and multi-file debugging. Safety improvements include real-time misuse safeguards, a 0.09 % misaligned action rate and 0 % refusal on Terminal-Bench-Science, while maintaining Gemini 4 Carbon-level protections against prompt injection. Neon’s “xhigh” thinking level delivers deeper reasoning with 36 % lower cost per task than competing frontier models at their highest settings.

---

*Original Think-Tank Source: [Introducing Gemini 4 Neon](https://deepmind.google/blog/introducing-gemini-4-neon/)*
