Categories
Technology

Google unveils Gemini 4 Neon

Google has introduced Gemini 4 Neon, a new artificial intelligence model aimed at businesses and developers looking for advanced AI capabilities without the high costs associated with powerful computing models. The launch adds to the growing competition among Google, OpenAI and Anthropic as technology companies race to build AI systems that can do more than answer questions and generate content.

The latest model is designed to handle complex assignments involving data analysis, software development, scientific research and business automation. Google is positioning Gemini 4 Neon as a middle-ground option that offers a combination of intelligence, speed and affordability, making it suitable for organisations that need AI support for regular, demanding workloads.

Gemini 4 Neon occupies a place between Google’s high-performance Gemini 4 Carbon and the faster, lighter Gemini 4 Hydrogen. While Carbon is built for complicated tasks that require extensive reasoning, Hydrogen is intended for simpler assignments where quick responses are the priority. Neon seeks to bring the strengths of both into a single model, giving businesses a more balanced option.

A major focus of the new model is its ability to carry out multi-step tasks with limited human intervention. Rather than simply responding to instructions, it can break down assignments, work with different tools, analyse information and produce finished results. This makes it particularly relevant to companies exploring AI agents to automate routine processes and improve workplace productivity.

For businesses, the applications could range from examining financial records and preparing reports to analysing large datasets and developing software. The model is also designed to assist programmers with understanding complicated codebases, identifying errors and working through technical problems.

Another notable feature is its one-million-token context window. This allows Gemini 4 Neon to process large volumes of information in a single session, including lengthy documents, spreadsheets, presentations and other forms of content. The capability could be particularly useful for organisations that regularly work with extensive records and need to identify connections across multiple sources.

Google is also highlighting the model’s ability to work with spreadsheets and perform complicated calculations. In a demonstration, the company showed how multiple AI agents using Gemini 4 Neon recreated an image through a spreadsheet containing approximately 100,000 cells and more than 500,000 formulas. The exercise illustrated the model’s ability to use spreadsheet functions creatively to complete a complex task.

The emphasis on such capabilities reflects a broader change in the AI industry. Companies are increasingly looking for systems that can take responsibility for completing assignments rather than simply offering suggestions or generating text.

How Gemini 4 Neon compares with rivals

Google is pitching Gemini 4 Neon against leading models from its competitors, including Anthropic’s Claude Opus 5.5 and OpenAI’s GPT-6.2 Sol. Its published benchmark results suggest that the new model is competitive across several areas, particularly business automation, reasoning and scientific problem-solving.

On the Artificial Analysis Intelligence Index, Gemini 4 Neon scored 61, compared with 59 for Claude Opus 5.5 and 57 for GPT-6.2 Sol. It also recorded 70.9% on the Vals Index, slightly ahead of Claude at 70.1% and OpenAI’s model at 66.1%.

The differences were more pronounced in certain business-focused evaluations. Gemini 4 Neon achieved 56.3% on AutomationBench, while Claude Opus 5.5 scored 42.5% and GPT-6.2 Sol recorded 41.3%. The benchmark measures how effectively AI models handle automated workplace tasks.

In scientific and mathematical reasoning, Neon scored 84% on RiemannBench, compared with 80% for both competing models. It also achieved 71.3% on HLE-Diamond, an evaluation designed to test advanced reasoning and knowledge.

However, Google does not lead across every category. Claude Opus 5.5 performed better in certain coding assessments, including Terminal-Bench 4.1, while GPT-6.2 Sol recorded a higher score in interpreting complex professional charts. The results are based on published benchmark comparisons and may vary depending on the evaluation methods.

Lower costs and wider accessibility

Pricing could be an important factor in Google’s attempt to attract enterprise customers. Gemini 4 Neon costs $3 per million input tokens and $15 per million output tokens. Cached input tokens are priced at just $0.15 per million, making repeated processing of information considerably cheaper.

Google has also claimed that Neon delivered comparable or better overall quality than Gemini 4 Carbon in its early evaluations, while running three times faster and at 60% of the cost. These improvements could make advanced AI applications more accessible to businesses that have been cautious about the expense of deploying powerful models at scale.

The model is being made available through the Gemini API on Google AI Studio, Gemini for business and Google’s AI-assisted development platform, Antigravity. Eligible Google AI Pro and Ultra subscribers can also access it through the Gemini app.

With Gemini 4 Neon, Google is attempting to address one of the biggest challenges facing enterprise AI adoption: delivering sophisticated capabilities without making deployment prohibitively expensive. As competition intensifies, the ability of AI models to complete real-world tasks accurately, quickly and economically could become as important as their reasoning capabilities.