What's New in GPT-6 Sol and Luna
GPT-6 Sol and Luna sit below GPT-6 Astra, OpenAI's top-tier model released earlier this month, rounding out a three-tier lineup. There is no GPT-6 Terra at this time. Sol is positioned for developers and knowledge workers handling repeated, complex work — building features, reviewing code, debugging, and analyzing data. Luna is aimed at simpler, high-volume tasks such as summarizing documents, extracting information, or answering quick questions at scale.
Both models were trained using methods similar to Astra's, bringing its underlying architectural advances down into faster, cheaper tiers rather than launching entirely separate model families. OpenAI attributes the price cuts specifically to gains in inference efficiency and prompt caching, rather than to any reduction in model quality or capability.
Pricing Breakdown
GPT-6 Sol: $2 per million input tokens / $10 per million output tokens (down from $4/$20 for GPT-5.6 Sol)
GPT-6 Luna: $0.10 per million input tokens / $0.50 per million output tokens (down from $0.20/$1.20)
An OpenAI spokesperson confirmed to VentureBeat that these are default, permanent prices — a notable shift from GPT-5.6, whose current rates were always meant to be promotional rather than standard.
Performance Claims
OpenAI reports that Sol makes roughly half as many factual errors as GPT-5.6 Sol on an internal test built from de-identified ChatGPT conversations in which users flagged an earlier model's mistake — though OpenAI itself notes this test set is deliberately skewed toward error-prone conversations and doesn't represent typical use. On computer-use benchmarks (OSWorld 2.0, offline), Sol at its highest effort setting scored 60.5%, narrowly ahead of Anthropic's Opus 5 at medium effort (60.3%), while costing roughly 80% less per task. On AutomationBench 1.0.6, Sol at high effort scored 33.2% at a cost of $0.27 per task. Luna, OpenAI says, can match GPT-5.6 Sol's performance at roughly one-hundredth of the cost when run at higher effort settings. Both new models reportedly also score better than their predecessors on OpenAI's internal alignment evaluations, including a lower rate of misleading claims about their own coding work.
Where to Access Them
GPT-6 Sol and Luna are live now in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu subscribers, with Luna also available to Free and Go users through the desktop app. Developers can call both models directly via the OpenAI API using the identifiers gpt-6-sol and gpt-6-luna. Neither model has reached the standard ChatGPT app yet — OpenAI says that rollout is happening gradually throughout the day of launch.
The Bigger Picture
This release didn't happen in isolation. Anthropic shipped Claude Opus 5.5 about 90 minutes before OpenAI's announcement, and xAI launched Grok 4.7 within the same two-day stretch — three major labs updating their lineups almost simultaneously. Industry observers are increasingly framing this less as a benchmark race and more as a price war, with Sol and Luna specifically positioned as the models OpenAI expects developers to actually build products on, rather than models meant to chase leaderboard rankings. Google's Gemini reportedly remains cheaper than GPT-6 Sol even after Gemini 3.8 Flash's scheduled price increase, keeping competitive pressure alive. More updates are expected soon on both sides — Anthropic has Sonnet 5.5 and Haiku 5.5 in the pipeline, and xAI has signaled Grok 5 will aim squarely at the frontier — with OpenAI's own DevDay on September 29 likely to add another chapter to the story.
