GPT-6 Astra is Open AI's flagship model for demanding end-to-end work. It is suited for advanced analysis, software engineering, deep research, scientific work, and document creation, with particular strengths in long-horizon agentic tasks that involve computer and browser use.

Open Router hosts the model via multiple providers, routing requests based on modes like Balanced, Nitro, or Exacto. Pricing varies by provider: Open AI and Azure list $10 input / $50 output per 1M tokens, while Open AI Flex charges $25 input, and Open AI Fast $20 input / $100 output. Average paid prices often drop due to caching and discounts.

Performance metrics show best latency at 1.87s and throughput up to 51 tokens per second. Uptime over 3 days is 100%, with availability at 99.35%. Open Router's routing improves availability to 98.97% versus 95.39% without it.

Top traffic comes from apps like 1.omp (418M tokens), Codex, Descript, Hermes Agent, and Claude Code. The model supports a 1M context and was released Sep 4, 2026.