GPT-6 Astra

GPT-6 Astra (3 September 2026) is OpenAI's flagship and the first of the GPT-6 line — built for end-to-end computer use, browser work, coding and research, with a 1,050,000-token context and five reasoning-effort levels up to max.

Best for

• Long, multi-step agentic work — filling forms, driving spreadsheets, operating a browser end to end
• Complex reasoning and software engineering across whole repositories
• Very long context: 922,000 tokens of input in a single request
• Document, slide and spreadsheet creation from a brief

How it compares — and the India angle

• At $10/$50 per 1M tokens it is twice Claude Opus 5's price, and its cached input ($1.00/1M) is four times Claude Fable 5.1's $0.25 — on agentic work that re-reads context all day, that one line decides the bill.
• Indian devs: the API default effort is low, not the level the marketing shows. Set reasoning.effort explicitly, and treat xhigh/max as a budget decision rather than a default.
• Watch the 272,000-token line: a request past it is billed at 2x input and 1.5x output for the whole request, not just the excess — chunk long documents rather than paying twice for all of it.
• It is the first OpenAI model to meet the company's critical cybersecurity capability threshold, which is why access began with vetted customers rather than everyone.