OpenAI’s GPT-6 Astra targets complex business work with stronger agents

OpenAI says GPT-6 Astra is now available in ChatGPT Work, Codex and its API as a model designed for demanding professional tasks. The company positions it as a system that can operate inside familiar applications, including software without an API, reducing the need for custom integrations or extensive workflow preparation.

That computer-use capability is central to the pitch. OpenAI says Astra completed Financial Modeling World Cup challenges about four times faster than the winning human competitor. Early customer feedback also points to stronger document analysis, presentation creation, coding and design work. Hebbia reported that Astra followed presentation briefs 17% more faithfully than the next-best model and sourced claims to the correct document 19% more often.

Performance and pricing

For developers, the model’s cost and benchmark results may be more consequential than the headline capabilities. GPT-6 Astra starts at $10 per million input tokens and $50 per million output tokens. On Terminal-Bench 4.0, which evaluates complex terminal tasks such as engineering, configuration and data analysis, it scored 57.9%, compared with 37.3% for GPT-5.6 Sol and 55.8% for Claude Fable 5.1. OpenAI estimates its cost per task at roughly 9% below GPT-5.6 Sol and 63% below Claude Fable 5.1 on that test.

OpenAI also says Astra helped its engineers reduce Codex turn latency by 25 times after an allocator change, though peak memory use rose by about 30%. The broader message is that agent builders may need less scaffolding: the model is intended to make fewer retries, follow company templates more closely and handle longer workflows with better judgment.

Source: OpenAI News