OpenAI halves prices on its mid-tier models with GPT-6 Sol and Luna

GPT-6 Sol costs $2 per million input tokens, half the prior Sol price, and Luna drops to $0.10, with strong agent results reported at far lower cost per task.

OpenAI has added two models, GPT-6 Sol and GPT-6 Luna, to the GPT-6 family that opened with GPT-6 Astra. The company says they were trained with methods similar to Astra's and aimed at lower cost and faster speed. In the API, Sol is priced at $2 per million input tokens and $10 per million output tokens, down from $4 and $20 for GPT-5.6 Sol. Luna is $0.10 and $0.50, down from $0.20 and $1.20. The API names are gpt-6-sol and gpt-6-luna.

OpenAI reports that Sol scores 33.2% on AutomationBench at $0.27 per task, against 26.9% for Claude Opus 5, which it says costs 11.1x as much per task. On DeepSWE v1.1, Sol at max effort gets 68.8% versus 69.9% for Claude Fable 5, at about 80% lower cost per task. Luna scores 66.6% there. Prompt caching also improved, with 90% discounts on cached input reads. GitHub reports that this cut prompt tokens needing fresh processing by over 50% across billions of requests.

For knowledge work, the shift is in cost per completed task. When capable models handle agentic work at a fraction of the price, more people and small teams can afford to automate research, coding and office routines. Luna reaching Free and Go users in the desktop app widens access further. Sol and Luna are in ChatGPT Work and Codex for paid tiers, but not yet in Chat.

Caveats: OpenAI says these results come from its research environment or API and may differ from production ChatGPT. The Claude Fable 5.1 AutomationBench cost leaves out Opus 5 fallback costs, which arose on about 40% of tasks. The factuality claim, about half as many mistakes as the predecessor, rests on conversations where users flagged errors, which OpenAI says are not typical usage. The announcement now carries a note pointing to a newer model, GPT-6.1 Sol.

Source: Google News