Product Launches

OpenAI unleashes GPT-6.2 Astra, a $8M cybersecurity model, to the public

Share
OpenAI unleashes GPT-6.2 Astra, a $8M cybersecurity model, to the public

OpenAI has released GPT-6.2 Astra, a cybersecurity model with capabilities previously restricted to approved security teams, at a cost of $8 per million input tokens. This move follows a week of significant price cuts in the AI industry, led by Google.

TL;DR

  • OpenAI releases GPT-6.2 Astra, a cybersecurity model with capabilities previously restricted to vetted teams.
  • Google undercuts competitors with Gemini 4 Neon at $3 per million tokens, leading on legal and tax benchmarks.
  • AI price cuts enable broader adoption in financial workflows, from contract review to financial modeling.

What happened

OpenAI released GPT-6.2 Astra and GPT-6.2 Sol on November 17, both meeting the company's 'Critical' cybersecurity threshold. The models are now available to the public, with Astra priced at $8 per million input tokens, 20% less than its predecessor. Astra scored 62 on the Artificial Analysis Intelligence Index and outperformed Anthropic's Claude Fable 5.2 on legal and financial benchmarks at a lower cost.

Google released Gemini 4 Neon on November 18, priced at $3 per million input tokens, less than half the cost of OpenAI and Anthropic's top models. Gemini 4 Neon ranked first on legal and tax benchmarks and is designed for accounting and spreadsheet modeling. Google also released Gemini 4 Hydrogen on November 12, which completes tasks 2.5 times faster than Gemini 3.8 Flash.

Google and Microsoft launched decision models this week. Google's Gemini Decision 1 evaluates up to 1,000 questions in a single call at a fraction of the cost of Astra. Microsoft released Microsoft-Decision-1.1, which runs more than twice as fast as its predecessor and can decline to decide when confidence is low.

Why it matters

The release of GPT-6.2 Astra means that any organization, including potential attackers, can access advanced cybersecurity capabilities. Banks and other financial institutions must now assume that attackers have the same tools as their defenders.

The significant price cuts led by Google enable broader adoption of AI in financial workflows. Firms can now perform tasks like contract review, tax work, and financial modeling at a fraction of the previous cost.

The competitive landscape is heating up, with companies like Alibaba and Tencent also releasing models at lower costs. This price war is making advanced AI capabilities more accessible to a wider range of users.

Key facts

  • GPT-6.2 Astra is priced at $8 per million input tokens, 20% less than its predecessor.
  • Astra scored 62 on the Artificial Analysis Intelligence Index, compared to 65 for Anthropic's Claude Fable 5.2.
  • Gemini 4 Neon is priced at $3 per million input tokens, less than half the cost of OpenAI and Anthropic's top models.
  • Gemini 4 Neon ranked first on Vals AI's Legal Research Bench and Tax Agent Bench.
  • Gemini 4 Hydrogen completes tasks 2.5 times faster than Gemini 3.8 Flash.
  • Gemini Decision 1 evaluates up to 1,000 questions in a single call at roughly one-eightieth the price of Astra.
  • Microsoft-Decision-1.1 runs more than twice as fast as its predecessor and can decline to decide when confidence is low.
  • Alibaba's Qwen4-Max reduced average cost per task by roughly 60% compared to its predecessor.

Context

This release follows OpenAI's initial restriction of GPT-6 Cyber to approved security teams two weeks ago, citing high risks for general release. The company has now introduced a monitoring system called Control Room to manage the risks associated with public access.

The AI industry is experiencing a significant price war, with companies like Google, Microsoft, Alibaba, and Tencent all releasing models at lower costs. This trend is making advanced AI capabilities more accessible to a wider range of users, particularly in the financial sector.

The release of decision models by Google and Microsoft highlights the growing importance of AI in automating routine tasks and improving efficiency in various industries.

Topics

Related coverage

Join the discussion

Have a take on this story? Weigh in with our community on Facebook.

💬 Discuss on Facebook →