News in Short

• Google has launched Gemini 4 Argon globally for developers and enterprises.
• The AI model targets coding, research, writing and complex professional tasks.
• Introductory API pricing starts at $2 per million input tokens.
• Google claims strong benchmark results and improvements in coding workflows.
• Introductory pricing ends on November 30, 2026.

Google has launched Gemini 4 Argon globally, targeting developers and enterprises that need AI for complex professional work. The new model focuses on coding, research, writing and tasks across specialised fields. These include finance, legal services, cybersecurity and science.

Google says Gemini 4 Argon outperforms competing frontier AI models on several benchmarks. It also claims the model secured the top position on the Vals Index, which evaluates AI performance across economically significant sectors.

However, benchmark rankings alone do not establish how well a model will perform in every real-world workflow. Its usefulness will also depend on task complexity, accuracy, cost and integration with existing tools.

Google Gemini 4 Argon availability and access

Developers and enterprise customers worldwide can access Gemini 4 Argon through the Gemini API. Google AI subscribers can also use it through Google Antigravity, the Gemini app and AI Mode in Search.

This wider availability gives businesses and individual users different ways to access the model. Developers can integrate its capabilities into applications, while subscribers can use supported Google products for everyday tasks.

Google is also expanding its Fairwind Program to hundreds of additional partners. The programme could help more organisations evaluate the model for specialised business applications.

Gemini 4 Argon pricing starts at $2 per million input tokens

Google has introduced promotional API pricing for Gemini 4 Argon. During this period, input tokens cost $2 per million, while output tokens cost $10 per million.

Google is also offering a 95% discount on cached input tokens. This could reduce costs for applications that repeatedly process the same context.

The introductory rates remain available until November 30, 2026. After that, input pricing will rise to $4 per million tokens, while output pricing will increase to $20 per million tokens.

For developers, the final cost will depend on how much text the model processes and generates. Therefore, applications with long prompts or frequent responses will need careful cost management.

Google highlights coding and engineering improvements

Google says Gemini 4 Argon can handle demanding engineering tasks and reason across large codebases. Its work with the Chrome team offers one example of these capabilities.

According to Google, Argon helped separate core webpage handling from a legacy folder containing around six million lines of browser features. That folder sits within Chromium’s broader codebase of approximately 40 million lines. Google says the work, which would normally take months, was completed in weeks.

Google also reports that teams using Argon in Antigravity increased agentic code submissions by more than 35%. The company says code rollbacks fell by nearly a third.

These figures suggest potential gains for software teams. However, they represent Google’s reported results rather than an independent assessment of performance across organisations.

AI model targets research, science and enterprise workflows

Gemini 4 Argon is not limited to software development. Google says enterprise customers and startups have tested it for deep research, document synthesis, long-term data analysis and reasoning across large codebases.

The companies and organisations named by Google include Deloitte, LexisNexis, Ambience Healthcare and Isomorphic Labs. Their reported use cases span professional research, document-heavy work and scientific analysis.

Google also says Argon optimised its Agent Development Kit. The changes reportedly reduced in-memory session preparation time by 40% and improved cold imports by 21%.

In scientific computing, Google claims the model optimised the open-source xTB quantum chemistry package. The resulting molecular structure predictions were up to 2.3 times faster, according to the company.

Google further reports a 2.85-times inference speedup for Gemma 4 E2B running on an A100 GPU. These results highlight the model’s potential beyond general-purpose chat and content generation.

What Gemini 4 Argon means for developers

Gemini 4 Argon’s launch reflects Google’s push to make AI more useful for complex, multi-step work. Its focus extends beyond answering questions to supporting software engineering, research and specialised business processes.

The promotional API rates may encourage developers to test the model before pricing increases. Still, teams should evaluate output quality, reliability, latency and total usage costs against their specific needs.

Google’s reported results offer an early indication of the model’s intended strengths. Independent testing and wider deployment will provide a clearer picture of how consistently Gemini 4 Argon delivers those gains.