OpenAI launches GPT‑6 Sol and Luna on ChatGPT, Codex and API
OpenAI said Sol makes about half as many mistakes as its predecessor, and the company's pricing page lists Luna at a twentieth of Sol's short-context rate.
OpenAI launched two models on Tuesday, GPT-6 Sol and GPT-6 Luna, which it said are cut from the same cloth as GPT-6 Astra. Sol is available in ChatGPT Work, Codex and the API, TechCrunch reported.
Luna reaches those surfaces plus the desktop app, and Free and Go users, according to TechCrunch. OpenAI's pricing page lists Sol at $2 per million input tokens and $10 per million output tokens on short context, rising to $4 and $15 on long context.
Luna is listed at $0.10 per million input tokens and $0.50 per million output on short context, and $0.20 and $0.75 on long context. Accounts tracking the rollout posted the short-context figures alone, which understates what a long-context request costs.
OpenAI said Sol makes about half as many mistakes as its predecessor and reaches Astra-level reliability at lower cost. The figure comes from an internal factuality evaluation built on de-identified real-world conversations, according to TechCrunch. It is the company's own measurement.
Several accounts said the models appeared in Codex before any announcement. Testing Catalog and wallstengine both posted that the two models were rolling out. Nobody has published independent benchmark scores, and a 1.05-million-token context figure posted by wallstengine could not be verified.