Google DeepMind Unveils Gemini 4 Argon, a Frontier AI Model
New AI model sets benchmarks in coding, cybersecurity, and enterprise workflows.
1Frontier AI capabilities
Gemini 4 Argon represents a significant leap in AI performance, particularly in complex workflows like software engineering, legal and finance tasks, and cybersecurity defense. The model is currently being rolled out to trusted cyber defenders through Google's Fairwind Program. Its capabilities are already being utilized internally by thousands of Google employees for specialized coding tasks and research.
2Pricing and token limits
The model will launch at an introductory price of $2 per million input tokens and $10 per million output tokens, with cached input tokens priced at 95% off. Google is significantly expanding the model's output token limit to an industry-leading 1 million tokens, up from 64K, enabling deeper reasoning and problem-solving capabilities for complex, long-horizon tasks.
3Enterprise workflow transformation
Gemini 4 Argon is demonstrating exceptional performance across various enterprise applications. It sets new standards on benchmarks like DeepSWE v1.1 (77.9%) for software engineering tasks and leads on the Vals Index measuring economic impact across multiple sectors. The model also excels in domain-specific evaluations like financial research and legal drafting.
4Cybersecurity breakthrough
Trained specifically for cybersecurity defense, Argon can autonomously find, validate, and patch critical software vulnerabilities. Wiz is already using it through its Scan for Good initiative, where it uncovered a critical vulnerability in global healthcare software that previous models had missed. The model ties for first place on CWE-bench v1 with a top score of 68%.
5Visual understanding advancements
Gemini 4 Argon shows unique strengths in tasks requiring visual understanding, capable of professional chart analysis and long video comprehension. It achieves state-of-the-art performance on LVBench for long video understanding with a score of 91.7%. This capability extends its usefulness across various professional domains that rely on visual data interpretation.
6Safety and safeguards
Google is implementing robust safeguards across four key areas: misuse prevention, prompt injection defense, misalignment monitoring, and system hardening. The model is designed to refuse harmful requests while preserving legitimate research, and it leads in prompt injection robustness on the Gray Swan's IPI benchmark. These measures are undergoing rigorous testing by internal and external red teams.
7Internal adoption and impact
Thousands of Google employees are already using Gemini 4 Argon for daily tasks ranging from debugging to large-scale codebase migrations. The model is helping teams accelerate engineering productivity and achieve breakthroughs. Its ability to sustain long, multi-step tasks is enabling new levels of productivity in complex workflows across the organization.
8Future roadmap and availability
After the introductory period, pricing will increase to $4 per 1M input tokens and $20 per 1M output tokens. Google plans to gradually expand access following government pre-release processes and early tester feedback. The company is committed to sharing security best practices with partners to improve ecosystem-wide safety as frontier AI capabilities advance.
Original article: Hacker News — Google DeepMind Unveils Gemini 4 Argon, a Frontier AI Model