Introducing Google Gemini 4 Neon: Efficient Performance for Complex Workflows
Following the recent launch of its predecessors, Carbon and Hydrogen, Google has released Gemini 4 Neon, designed to provide advanced reasoning capabilities and improved performance for complex, high-frequency tasks. This new model is celebrated for delivering results more quickly and at a lower cost compared to previous iterations.
Evolution of the Gemini Model Family
Prior to the release of Neon, the Gemini 4 family included:
- Hydrogen: Focused on everyday assistance.
- Argon: Tailored for code development and analysis, ideal for most software engineering tasks.
- Carbon: Designed for the most complex challenges, pushing the limits of model intelligence.
Neon is positioned between Hydrogen and Carbon, boasting performance that matches or exceeds that of Carbon in areas such as coding and analytical tasks while operating at enhanced efficiency.
Features of Gemini 4 Neon
Neon is now available to subscribers of the Gemini app, specifically those using the AI Pro and AI Ultra tiers. The app has been updated to include:
- Adaptive Model Picker: Automatically adjusts to the user’s goals.
- Everyday Help: Hydrogen model for routine tasks.
- Enhanced Analysis: Neon for deeper insights.
- Complex Work: Carbon model for advanced operations.
Anticipated Outcomes with Neon
Google anticipates that users will benefit from:
- Improved quality in writing and knowledge work, delivering coherent and well-organized responses.
- Reliable research capabilities, integrating information from various web sources for timely updates.
- Accelerated processing for complicated tasks, allowing the completion of multi-step processes more efficiently than before.
The model maintains a 1 million-token context window, similar to Hydrogen, and features four distinct levels of cognitive processing: Low, Medium, High, and Extra High.
Performance Benchmarks
Benchmarks for Neon demonstrate its competitive edge:
- Ranked #3 overall with a score of 61 on the Artificial Analysis Intelligence Index.
- #2 in steerability on the Agent Arena Index, reflecting adaptability to user corrections.
- Top performance in specialized areas such as legal and tax research.
- Achieved #1 rankings in various financial and analytical categories on the Surge AI platform.
- Excelled at reading complex professional charts, scoring 73.5% on Chartography.
- Delivered strong results across diverse benchmarks, including coding and simulation tasks.
Pricing Structure
The pricing for Neon is set at $3 per 1 million input tokens and $15 per million output tokens, making it notably more expensive than Hydrogen. However, promotional pricing for the other models (Argon and Carbon) remains lower until November 30:
- $2 per 1 million input tokens and $10 per million output tokens.
- Post-promotion, the cost will rise to $4 per million input tokens and $20 per million output tokens for all models.
Overall, Neon offers a cost-effective alternative while providing enhanced capabilities compared to its predecessors.
