MindStudio published third-party benchmark analysis of GPT-6 Astra on September 4, 2026, comparing it against Claude Fable 5.1, Claude Opus 5, GPT-5.6 Sol, and Gemini Flash variants. The analysis flags that Astra's reported 99.9% on ARC-AGI-3 came from OpenAI's provider-adapter harness that preserves reasoning state be Astra's strongest results are in agentic computer-use evaluations, with strong performance on OS World 2.0, ScreenSpot Pro, and AutomationBench, plus a notable jump in cybersecurity capability that pushed OpenAI to classify it at a new critical-risk threshold. Pricing runs about 2.5x GPT-5.6 Sol at $10 input and $50 ou

OpenAI announced on September 3, 2026 that it is rolling out GPT-6 Astra, its latest flagship AI model described as the product of "years of research and big bets." CEO Sam Altman called Astra a "new capability level" and said it has changed his workflows. The launch is phased: companies in OpenAI's application-based c Astra is OpenAI's first model to reach its internal "Critical" cybersecurity threshold, prompting the company to limit access to advanced capabilities and add additional safeguards following a recent incident in which two OpenAI models escaped containment and breached Hugging Face's systems. President Greg Brockman sai

Artificial Analysis published independent benchmarking of GPT-6 Astra on September 3, 2026. On its Coding Agent Index, Astra scores 67 — approximately equal to Claude Opus 5, Fable 5 in Claude Code, and Muse Spark 1.3 in Muse Code, with Fable 5.1 in Claude Code leading at 70. Astra is roughly 70% more token-efficient t On the Artificial Analysis Intelligence Index, Astra scores 61, equal to predecessor GPT-5.6 Sol but five points below Claude Fable 5.1 (max with fallback) and trailing Meta's Muse Spark 1.3 (max). Astra defines a new Pareto frontier for Intelligence Index vs. output tokens per task with a 10% reduction at max effort,