GPT-6 Astra
OpenAI's GPT-6 flagship β a generational leap in computer use, software engineering, science, and professional work that reaches the Critical cybersecurity threshold under OpenAI's Preparedness Framework
GPT-6 Astra
OpenAI β’ September 2026
Training Data
Not disclosed
GPT-6 Astra
September 2026
Parameters
Not disclosed
Training Method
Reinforcement Learning with Chain-of-Thought Reasoning
Context Window
1,050,000 tokens
Knowledge Cutoff
April 30, 2026
Key Features
1M Token Context β’ Autonomous Computer Use β’ Critical-Tier Cybersecurity β’ Reasoning Effort: Low to Max
Capabilities
Computer Use: Maximum (72.6% OSWorld 2.0)
Science & Math: Maximum (96% GPQA Diamond)
Alignment: Outstanding (91.5% cyber-jailbreak refusal)
What's New in This Version
Beats GPT-5.6 Sol on OSWorld 2.0 (72.6% vs 65.7%) at ~47% less time per task, saturates ExploitBench (100%) and FrontierMath Tier 4 (~98%), and refuses 91.5% of cyber jailbreaks vs Sol's 59%; priced as a premium tier ($10/$50 per 1M tokens, fast mode $20/$100); rolling out in phases from September 3, 2026, starting with gated enterprise access
OpenAI's GPT-6 flagship β a generational leap in computer use, software engineering, science, and professional work that reaches the Critical cybersecurity threshold under OpenAI's Preparedness Framework
What's New in This Version
Beats GPT-5.6 Sol on OSWorld 2.0 (72.6% vs 65.7%) at ~47% less time per task, saturates ExploitBench (100%) and FrontierMath Tier 4 (~98%), and refuses 91.5% of cyber jailbreaks vs Sol's 59%; priced as a premium tier ($10/$50 per 1M tokens, fast mode $20/$100); rolling out in phases from September 3, 2026, starting with gated enterprise access
Technical Specifications
Key Features
Capabilities
Other OpenAI Models
Explore more models from OpenAI
GPT-5.6 Sol
OpenAI's flagship GPT-5.6 model β its most capable system yet for complex coding, cybersecurity, and scientific reasoning, with new max reasoning effort and ultra subagent mode
GPT-5.6 Terra
The balanced mid-tier of the GPT-5.6 family β matches GPT-5.5 quality on high-volume business tasks at roughly half the price
GPT-5.6 Luna
The fastest, most cost-efficient GPT-5.6 model β strong everyday capability at OpenAI's lowest price for summarization, drafting, and routine automation
