OpenAI has introduced two new GPT-6 models today, named Sol and Luna, expanding its portfolio beyond GPT-6 Astra. The company says these models bring Astra’s improvements to more affordable tiers, offering advancements in factuality, coding, general reasoning, and alignment. Sol is designed for higher reasoning capability, while Luna prioritizes fast responses and handling high-volume usage.
Performance Gains and Testing
According to internal benchmark tests, GPT-6 Sol cuts the error rate in half compared to its predecessor, GPT-5.6 Sol. It also delivers stronger results in coding and desktop workflows, matching or exceeding scores set by Anthropic’s Claude Fable 5.1 on some benchmarks. Meanwhile, OpenAI also evaluated GPT-6 Sol against Opus 5, but Anthropic launched Opus 5.5 today, which outperforms Astra in some coding and knowledge domains.
Pricing, Efficiency, and Access
Prices for input and output tokens have been significantly lowered. GPT-6 Sol now costs $2 per million input tokens and $10 per million output tokens—roughly half the price of Sol 5.6. Luna is priced at $0.10 per million input tokens and $0.50 per million output tokens, down from previous rates of $0.20 and $1.20 respectively. Also improved is prompt caching: agents now reuse more context, respond faster, and get up to 90% discount on cached input-token reads.
In terms of availability, both Sol and Luna are rolling out in ChatGPT Work and Codex starting today for Plus, Pro, Business, Enterprise, and Edu users. Luna is additionally accessible to Free and Go users through the desktop app. Neither model is yet supported in the Chat mode.
OpenAI says both models adopt Astra’s new style of communication: more concise, less jargon, balanced outputs with fewer misleading statements, especially around coding. Responses are intended to keep substance while cutting down low-value detail.
What this means:With Sol and Luna, OpenAI seems committed to broadening the practical accessibility of its most recent architecture. The pricing moves suggest that the company views token cost as a barrier to adoption. If Sol delivers on its promise of reducing errors and improving coding fluency, it could shift many users away from higher-cost, higher-latency models. Luna might become the go-to for applications needing speed and bulk processing.
