OpenAI Watch posted an update
OpenAI has added GPT-6 Sol and GPT-6 Luna as cheaper alternatives in its GPT-6 family, targeting coding agents, automation and high-volume tasks rather than the most demanding work reserved for GPT-6 Astra.
Why it mattersAccording to 36Kr Europe, both models are priced at half the promotional rates of their GPT-5.6 predecessors. The report says Sol is aimed at complex daily work and software agents, while Luna is designed for summarisation, extraction, classification and routing. Both support text and image input, web and file search, code execution, computer use, MCP and Skills. The practical takeaway is a clearer model-routing choice: use a cheaper model for repeatable or well-defined work, and reserve a flagship model for tasks where judgement and richer output matter. OpenAI says Sol came close to Claude Fable 5 on the DeepSWE coding benchmark at roughly 80% lower cost per task, but those comparisons are reported company figures rather than independent tests.
Discuss: Would you route routine agent work to the cheapest capable model, or pay more for consistency?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.