Skip to content
The Daily Circuit
AI News

Google DeepMind Launches Gemini 4 Neon Model With 1 Million Token Context Window

Google DeepMind has introduced Gemini 4 Neon, a new model offering frontier-level intelligence, a 1 million-token context window, and advanced capabilities for software engineering, knowledge work, and science workflows.

By The Daily Circuit AI Desk (automated, AI-generated)

2 min read

Abstract digital illustration representing AI model architecture and data processing flows.

AI-generated image (flux-1-schnell), not a photograph of the events described.

Key facts

  • Gemini 4 Neon features a 1 million-token context window.[1]
  • The model is priced at $3 per million input tokens, $0.15 per million cached input tokens, and $15 per million output tokens.[1]
  • Neon ranks #3 overall on the Artificial Analysis Intelligence Index with a score of 61 at its xhigh thinking level.[1]
  • Neon achieves #1 on the Legal Research Bench and Tax Agent Bench.[1]
  • Gemini 4 Neon for cyber is rolling out to trusted defenders via the Fairwind Program.[1]

Google DeepMind has announced the launch of Gemini 4 Neon, a model designed to deliver frontier-level intelligence with fast end-to-end task completion across complex software engineering, knowledge work, and science workflows [1]. The model is rolling out across the Gemini API in Google AI Studio, Google Antigravity, and the Gemini app [1].

Model Performance and Capabilities

Gemini 4 Neon ranks #3 overall with a score of 61 on the Artificial Analysis Intelligence Index at its xhigh thinking level, and takes the #2 spot on the Vals Index landing on the Pareto frontier [1]. It also ranks #2 on steerability on the Agent Arena Index [1]. In enterprise, analytical, and scientific domains, Neon handles inputs across spreadsheets, PDFs, charts, audio, and video within a single 1 million-token context window [1].

Independent evaluations show Neon ranking #1 on the Legal Research Bench and Tax Agent Bench, as well as Surge AI benchmarks covering finance, accounting, spreadsheet modeling, and complex constraint following [1]. It achieves #1 on Zapier’s AutomationBench and Andon Labs' Vending-Bench 2, and ranks #2 globally on Arena's Agent Arena Index [1]. For scientific and mathematical reasoning, the model is best in class on Terminal-Bench Science, RiemannBench, and HLE-Diamond [1].

Developer and Coding Features

Engineered for long-running agent workflows, Neon sustains performance through tool-heavy tasks, either operating independently or acting as an Orchestrator directing parallel Hydrogen subagents in Google Antigravity [1]. It ranks #1 overall on the Artificial Analysis Coding Agent Index, and scores strongly on Terminal-Bench 4.1 and FrontierCode 2.0 [1]. On the mini-SWE harness, it reaches 12.8%, which increases to 39.0% within Google Antigravity [1].

Cyber Defense and Security

Makes headway in cyber defense, Gemini 4 Neon generates secure code across easy, medium, and hard challenges on AutoBaxBench, achieving a state-of-the-art score on the hard tier [1]. On WebExploitBench v1.2, it achieves 51.26% at half the cost of competing large models [1]. Furthermore, Neon records state-of-the-art performance on SEC-bench Pro, achieves frontier-level performance on ExploitGym, and reaches 70.7% on Collinear’s CWE-bench v2 [1]. Gemini 4 Neon for cyber is rolling out to trusted defenders through the Fairwind Program [1].

Pricing and Availability

The model is priced at $3 per million input tokens, with cached input tokens discounted by 95% to $0.15 per million, and output tokens priced at $15 per million, featuring a 1 million-token context window [1]. Neon is available immediately in the Gemini API in Google AI Studio, Gemini for business, paid Google AI Pro and Ultra subscribers in Google Antigravity, and Pro and Ultra subscribers in the Gemini app [1].

Context

Gemini 4 Neon expands Google DeepMind's model lineup, introducing refined thinking levels that scale from default high to xhigh settings to balance intelligence and token spend across analytical, scientific, and software engineering workloads [1].

Why it matters

The release provides enterprise and developer users in India and globally with a high-capacity, cost-efficient tool for long-running agentic workflows, automated coding, and cyber defense operations.

What could happen next

AI-generated analysis based on the sources below. Not a forecast or recommendation.

The introduction of cost-efficient, long-context models like Gemini 4 Neon could accelerate enterprise adoption of autonomous agentic workflows and advanced software engineering assistants.

Risks and open questions

As with any advanced AI model, open questions remain regarding real-world reliability in unconstrained environments and the ongoing mitigation of misuse and prompt injection risks.

Sources

  1. 1.Google DeepMind (primary source): Introducing Gemini 4 Neon, 19 November 2026

How this article was produced. This story was researched and written by an automated system using Google’s Gemini models, working only from the sources listed above. Automated checks verified that every citation, quote and figure matches those sources before publication. It was not written by a human reporter. Editorial policy · Report an error

  • Cybersecurity
  • Software Engineering
  • Artificial Intelligence
  • Google DeepMind
  • Gemini 4 Neon