Google Launches Gemini 4 Argon, AI Model for Cybersecurity

Key Takeaways

• Google launched Gemini 4 Argon, a frontier AI model focused on coding and cybersecurity, initially available only to trusted cyber defenders through its Fairwind Program.
• Argon leads on 14 of 19 benchmarks and scored 53 on the Artificial Analysis Intelligence Index, tying with GPT-6 Astra but trailing Anthropic's Claude Opus 5.5 at 58.
• The model can generate up to 1 million output tokens, far exceeding the 128K caps common among OpenAI and Anthropic models.
• Google priced Argon at $2 per million input tokens and $10 per million output tokens, making it about 40% cheaper per task than GPT-6 Astra at a 50% promotional discount.
• Google is withholding Argon from public release and sharing it with the U.S. government for early feedback, mirroring Anthropic's cautious approach.
• Alphabet shares rose about 2% after the announcement, while JPMorgan maintained an Overweight rating but urged Google to reclaim AI leadership.
• AI is accelerating vulnerability discovery: high-risk exploitation more than doubled in the first eight months of 2026, and AI agents find RCE flaws at twice the traditional rate.
• The US FTC opened an inquiry into Anthropic, OpenAI, and other AI labs focusing on security practices.
• Synopsys announced AI partnerships with OpenAI and Amazon Web Services, including a multi-year license agreement worth over $1 billion.
• The top 1% of companies spend about eight times more on AI vendors than the top 10%, with median spending of roughly $800,000 versus under $100,000.

Google launches Gemini 4 Argon AI model for cyber defenders

Google announced Gemini 4 Argon, a new frontier AI model, on Wednesday. It is initially rolling out to trusted cyber defenders through its Fairwind Program. Google says Argon can autonomously find, validate, and patch critical software vulnerabilities. The company will expand access gradually as it gathers feedback from early testers.

Google unveils Gemini 4 Argon with strong benchmark results

Google announced Gemini 4 Argon, its new frontier AI model, after more than half a year away from the AI race. Argon leads on 14 of 19 benchmarks and launches at an introductory price of $2 per million input tokens and $10 per million output tokens. The model is rolling out first to trusted cyber defenders through Google's Fairwind Program. Google says thousands of employees are already using it for coding, research, and writing tasks.

Gemini 4 Argon matches GPT-6 Astra score at lower cost

Google's Gemini 4 Argon scored 53 on the Artificial Analysis Intelligence Index, tying with GPT-6 Astra. At a 50% promotional discount, Argon costs about 40% less per task than Astra. Anthropic's Claude Opus 5.5 still holds the top score at 58. Argon also takes the top spot on AutomationBench-AA with 77.5% and has the lowest hallucination rate at 15% among models scoring above 45.

Google releases Gemini 4 Argon as its most powerful AI model

Google released Gemini 4 Argon on September 30, 2026, calling it its most powerful model yet. The model is being rolled out to a select group of cyber partners through its Fairwind Program. Google says Argon can autonomously find, validate, and patch critical software vulnerabilities. The company also highlights Argon's strengths in coding, engineering, and visual parsing.

Google limits public access to Gemini 4 Argon over safety

Google said it will withhold Gemini 4 Argon from the public and release it only to vetted cybersecurity experts. The company is voluntarily giving the U.S. government early access and will gather feedback before a wider release. Google says Argon excels at software engineering, legal and financial work, and cyber defense. The cautious rollout mirrors rival Anthropic's approach with its most advanced model.

Alphabet shares rise as Google launches Gemini 4 Argon AI model

Alphabet shares rose about 2% after Google launched Gemini 4 Argon, an AI model focused on coding and cybersecurity. The model will first go to trusted cybersecurity researchers, then to paid API customers and Google AI Ultra subscribers. Jefferies priced it at $2 per million input tokens and $10 per million output tokens. Google said Argon scored 77.9% on the DeepSWE v1.1 coding benchmark and can find and fix software vulnerabilities on its own.

Google launches Gemini 4 Argon for cybersecurity partners first

Google is releasing Gemini 4 Argon, its most powerful AI model, to cybersecurity partners before the public. The model can find, verify, and patch software vulnerabilities on its own. It has a 1 million token output limit, up from 64,000. Google priced it at $2 per million input tokens and $10 per million output tokens with a 95% discount on cached input tokens. The company has not set a date for a broad public release.

Google launches Gemini 4 Argon after months of delays

Google released Gemini 4 Argon after a seven-month wait, marking its first frontier AI model release in months. The model is currently only available to cybersecurity firms. Google plans broader availability once safety issues are fixed. The model performs well in coding, finance, and other tasks. Some employees were unhappy with the development process, but Google pushed the release forward.

GOOGL stock rises after Gemini 4 Argon launch

Alphabet shares rose after Google unveiled Gemini 4 Argon, a frontier AI model for software engineering and cybersecurity. JPMorgan kept an Overweight rating but said Google must reclaim AI leadership. The model competes with OpenAI and Anthropic models on coding and knowledge tasks. Google is releasing it in phases, starting with trusted cyber defenders through its Fairwind program. It launches at $2 per million input tokens and $10 per million output tokens.

When you may be able to use Google's new Gemini 4 Argon

Google announced Gemini 4 Argon, its newest frontier AI model, but it is not yet available to the public. Right now it is only available to cybersecurity defenders in Google's Fairwind program. The model can handle coding, images, charts, and reasoning. Google says it is using it internally for quantum computing research and memory operations. The company has not given a timeline for wider release.

Google releases powerful Gemini 4 Argon AI model to select experts only

Google released its most powerful AI model, Gemini 4 Argon, only to vetted cybersecurity experts. The company wants to prevent hackers from misusing the technology. Google is also sharing the model with the US government for early feedback before a wider release.

Gemini 4 Argon can output 1M tokens, far exceeding rival models

Gemini 4 Argon can generate up to 1 million output tokens in one response. Most leading models from OpenAI and Anthropic cap output at 128K tokens. This gives Argon a major edge for long coding tasks, legal work, and research, though pricing and quality over long outputs remain concerns.

Google DeepMind launches Gemini 4 Argon for coding and cyber defense

Google DeepMind unveiled Gemini 4 Argon, its first Gemini 4 generation model. It supports coding, legal and finance knowledge work, and cybersecurity defense. Argon leads on some benchmarks like DeepSWE v1.1 and the Vals Index but trails on others. Access is currently limited to trusted cyber defenders through the Fairwind Program.

AI agents discover RCE flaws twice as fast as attackers exploit them

AI research agents now find remote code execution vulnerabilities at a rate of 50%, double the traditional 26%. Hackers exploit these flaws within days, as seen in the CVE-2026-1731 case. Monthly vulnerability disclosures doubled in 2026, and high-risk findings surged by 167%.

Google GTIG reports AI speeding up vulnerability discovery and exploitation

Google Threat Intelligence Group found that AI is accelerating vulnerability discovery across enterprises. High-risk vulnerability exploitation more than doubled in the first eight months of 2026. AI-discovered flaws are more likely to be medium or high risk and lead to remote code execution.

Temasek stays positive on AI spending despite high costs

Singapore investor Temasek says it is not worried about huge AI spending by tech giants. Executive Jane Atherton spoke at a Reuters event in Boston on October 1, 2026. She said the companies funding AI infrastructure believe the investments will bring solid returns. Temasek owns stakes in Anthropic and OpenAI and has about $400 billion in total holdings.

Temasek confident in AI investment despite high capex

Temasek's North America Head Jane Atherton said the firm remains positive on AI. She spoke at the Reuters Investment USA event in Boston on October 1, 2026. Hyperscaler spending on AI is forecast to pass $1 trillion by 2027. Temasek plans to grow its AI portfolio share from about 6% to as much as 15% over five years. Its Americas holdings make up 26% of its global portfolio, including stakes in BlackRock, Mastercard and Nvidia.

Flytxt named Challenger in Gartner AI marketing report

Flytxt was recognized as a Challenger in the 2026 Gartner Magic Quadrant for CSP AI-Enabled Marketing and Sales Solutions. CEO Dr. Vinod Vasudevan said the company is building an AI Expertforce called Niya-X. Niya-X uses causal reasoning, counterfactual simulation and federated learning to help communications service providers. The system supports sales and marketing decisions across the value chain.

US FTC opens inquiry into Anthropic and OpenAI

The US Federal Trade Commission has opened an inquiry into Anthropic, OpenAI and other AI labs. The request for information was opened on Tuesday and focuses on security practices. The FTC has been increasing its scrutiny of the AI industry in recent years. A spokesperson for Anthropic said the company is cooperating with the FTC. The exact nature of the inquiry remains unclear.

Mynd.ai stock surges on heavy trading volume

Mynd.ai stock rose 21.39% to $0.37 in premarket trading on October 1, 2026. The micro-cap company saw volume jump to 5.48 million shares versus a 100-day average of about 30,700. In first-half 2026, Mynd.ai reported revenue of $73.4 million and a net loss of $20.2 million. The company also cut operating expenses 35% to $31.5 million. Separately, Promethean launched AI Companion for its Explain Everything app using Apple Intelligence.

Synopsys Sets Higher Revenue Targets and Secures Major AI Deals

Synopsys used its investor day in New York to set higher fiscal 2027 revenue targets. The company announced new AI partnerships with OpenAI and Amazon Web Services. A multi-year license agreement worth over $1 billion was revealed. Synopsys says the AI boom is increasing demand for its chip-design tools.

Top 1% Companies Spend Far More on AI Than Other Large Firms

A report by Andreessen Horowitz tracked enterprise AI spending through July 2026. The top 1% of companies now spend about eight times more on AI vendors than the top 10%. Median spending for the top 1% reached roughly $800,000, while the broader top 10% stayed below $100,000. The gap shows AI adoption is uneven across businesses.

Industry Leaders Map Out Future for AI Data Center Hardware

Three executives discussed the growing data center hardware market in a Q&A. Velocity Group CEO Jorge Xacur expects strong demand for at least 10 years. Zayo CTO Troy Lupe projects AI capacity demand could grow two to six times by 2030. Eaton's Mike Yelton says customers need ongoing upgrades for power and cooling systems.

Netlist Patent Case Could Block Micron Memory Imports for AI Hardware

Netlist filed a complaint with the U.S. International Trade Commission against Micron. The complaint alleges infringement of two high-bandwidth memory patents. An ITC case could result in products being excluded from the U.S. market. High-bandwidth memory is a key component in AI accelerators used by companies like Nvidia and Google.

Sources

• Google Launches Gemini 4 Argon With Guardrail-Free Access for Vetted Defenders
• Google Introduces Gemini 4 Argon, Beats GPT 6 Astra And Opus 5.5 On Many Benchmarks
• Gemini 4 Argon Ties With GPT-6-Astra On Artificial Analysis Intelligence Index At A 40% Cheaper Price
• Google releases Gemini 4 Argon, called its most powerful model yet
• Google restricts access to new AI model over safety concerns
• Alphabet Shares Rise After Google Launches Gemini 4 Argon AI Model
• Google is launching Gemini 4 Argon, its most powerful AI model, for cybersecurity partners first
• Google launches Gemini 4 Argon, its first AI frontier model release in months
• GOOGL Stock Rises After Gemini 4 Argon Launch, But JPMorgan Says Google Must Reclaim AI Leadership
• Gemini 4 Is Here: When You May Be Able to Use Google's New AI Model
• Google rolls out new Gemini AI model but restricts access over safety concerns
• Gemini 4 Argon Can Generate 1 M Output Tokens, Compared To 128k For Most Other Frontier Labs
• Google DeepMind Unveils Gemini 4 Argon with 1M Output Tokens for Coding, Knowledge Work and Cyber Defense
• AI Agents Find RCE Vulnerabilities at Double the Traditional Rate—And Attackers Exploit Them in Days
• Google GTIG finds AI accelerating vulnerability discovery across enterprise and critical infrastructure attack surfaces
• Temasek comfortable with current AI spending levels, executive says
• Temasek Confident in Current AI Investment Despite High Capex
• Flytxt Recognised as a Challenger in the 2026 Gartner® Magic Quadrant™ for CSP AI-Enabled Marketing and Sales Solutions
• US regulator opens inquiry into Anthropic, OpenAI and other AI labs
• Why Is Mynd.ai Stock Gaining Thursday?
• Synopsys Turned Its Investor Day Into An AI Sales Pitch
• AI Investment Gap Widens as Top 1% Companies Now Spend More on The Technology Than The Top 10%
• 3 Execs Outline Strategy for AI Data Center Hardware Boom
• Micron Import Ban Could Hit Nvidia and Google AI Hardware Supply