OpenAI tiered model lineup

The lineup is doing two jobs at once, it lets OpenAI charge premium prices for the hardest work while making routine usage cheap enough to spread everywhere. Astra sits at the top for complex reasoning, coding, research, and long context. Sol is the middle tier for strong performance at much lower cost. Luna is the fast, low cost default that makes everyday chat and lightweight agent work economical at massive scale.1 2 3 4

• The price gaps are large enough to shape product behavior. In the API, GPT-6 Astra is listed at $10 input and $50 output per 1M tokens, GPT-6.1 Sol at $2 and $10, and GPT-6 Luna at $0.10 and $0.50 in the ChatGPT enterprise rate card. That creates a clear ladder from premium reasoning to commodity throughput.2 3 4
• OpenAI is also narrowing the gap between middle and top tiers. GPT-6.1 Sol is positioned as near Astra level performance for a fraction of the price, with lower cached input pricing than GPT-6 Sol. That matters most for coding agents and recurring workflows that reuse context over and over.2 5 6
• The model ladder maps cleanly onto product packaging. OpenAI says GPT-6 Astra is its most capable model for demanding work, while GPT-6.1 Sol and Luna are available in ChatGPT Work and Codex, and Luna is described as the most cost efficient option. This supports broad distribution in ChatGPT while preserving higher monetization in pro and enterprise workflows.1 2 4 7

This is heading toward finer grained routing, where users buy one product but OpenAI silently decides when to spend Astra dollars, when Sol is enough, and when Luna can handle the job. As model quality converges and token prices keep falling, the advantage shifts from having one best model to operating the best cost stack across consumer chat, agents, and enterprise work.2 4 5 7