Two New Models, One Clear Message: AI Is Getting Cheaper
OpenAI expanded its GPT-6 lineup on Tuesday, launching GPT-6 Sol and GPT-6 Luna at prices roughly half of what their predecessors cost — and the timing was no accident. According to Reuters, the two models arrived roughly 90 minutes after Anthropic released Claude Opus 5.5, turning an ordinary model launch day into a full-blown AI price war.
The new arrivals round out a GPT-6 family that began with this month’s flagship Astra. GPT-6 Sol targets complex work like coding and comes in at $2 per million input tokens and $10 per million output tokens — down from $4/$20 for GPT-5.6 Sol. GPT-6 Luna, built for high-volume clerical tasks, costs just $0.10/$0.50 per million tokens, down from $0.20/$1.20 previously.
Cheaper, Faster, and Fewer Hallucinations
OpenAI credits caching and inference improvements for the steep price drops. The company also says Sol makes about half as many factual mistakes as its predecessor — a meaningful claim in a space where hallucinations remain the industry’s most stubborn problem.
The pricing strategy reveals how OpenAI is segmenting its market. Sol is aimed squarely at the developer sweet spot: complex coding, reasoning, and multi-step tasks where accuracy matters and customers will pay a premium over cheaper tiers. Luna, at a fraction of a cent per thousand tokens, goes after the high-volume end — document classification, data extraction, and other clerical workloads where cost per token decides whether an AI pipeline is viable at all.
TechCrunch reports that both models are available in ChatGPT Work, Codex, and the API. Even Free and Go users get a taste: Luna is accessible in the desktop app, a notable move toward giving casual users a shot at the newest generation of models without a paid tier.
90 Minutes That Changed the Story
The real drama, though, is the choreography. Anthropic’s Claude Opus 5.5 — itself about 40% cheaper than Opus 5 — landed just 90 minutes before OpenAI’s announcement. That makes Tuesday one of the rare days in tech where two leading AI labs shipped competing price cuts within the same afternoon.
Whether the near-simultaneous launches reflect leaked timelines, aggressive product scheduling, or simply a race both labs knew was on, the signal to developers is unambiguous: the two biggest names in AI are now competing directly on price, not just capability.
Why This Matters
For developers and businesses, the math is simple: token costs for serious models just fell by roughly half in a single day. Workloads that were borderline on price — bulk document processing with Luna, heavy coding workloads with Sol — suddenly look a lot more attractive.
For the industry, Tuesday’s dual launch confirms that the frontier-model business has entered a new phase. Capability gains still matter, but cost per token is becoming the battleground that decides who wins enterprise deals. Expect the other labs to answer — Google, xAI, and DeepSeek rarely let a price move go unanswered for long.
There is also a defensive logic at play. Both OpenAI and Anthropic know that today’s developers are far less locked in than they were a year ago, with abstraction layers making it trivial to swap models behind an API. Cutting prices in half is one way to keep workloads from migrating — and to force smaller rivals, who can’t match these margins, further downmarket.
Meanwhile in AI Tech: DeepSeek, OpenAI to Brief UN Security Council on AI Risks
Meanwhile in Tech News: AMD Hits $1 Trillion Market Cap as AI Chip Rally Accelerates
