OpenAI slashes GPT-5.6 Luna price by 80% amid US rollout limits

OpenAI announced that its budget‑friendly GPT‑5.6 Luna model will cost $1 per million input tokens and $6 per million output tokens, an 80% discount versus the flagship Sol pricing of $5/$30 [3]. The move is aimed at igniting a token‑price battle even as the U.S. government restricts the models to a small, hand‑picked group of companies.

Pricing strategy and government constraints

OpenAI’s pricing sheet positions Luna as the cheapest tier, with Terra at roughly half the Sol rate and Sol itself at $5 / $30 per million tokens—about half the cost of Anthropic’s Claude Fable 5 ($10 / $50) and well below the rates of competing models [1][3]. The company framed the pricing as a push to make “powerful AI accessible without breaking the bank,” while simultaneously acknowledging a “tight leash” from the Trump administration that limits access to a curated set of partners [1][2].

The limited preview stems from a recent executive order that obliges AI developers to provide cutting‑edge models for federal assessment before broader release [2]. OpenAI agreed to a “small group of trusted partners” rollout in June, expanding to a public launch on July 9 after the government lifted the initial restriction [2]. The company warned that prolonged “government babysitting” could hamper developers and enterprises that need the tools now [1].

Competitive landscape and price‑war implications

The aggressive pricing directly challenges Anthropic, whose Claude Fable 5 still commands $10 / $50 per million tokens, and Meta’s newly launched Muse Spark 1.1 at $1.25 / $4.25 per million tokens [3]. While Meta’s rates are lower on the input side, OpenAI’s Luna remains the cheapest for output tokens, a key cost driver for agentic workflows that often generate large answer volumes. Analysts note that lower token rates do not automatically translate into lower total spend, as enterprise agents can consume many more tokens per task than simple chat interactions [3].

The price cuts echo a broader cloud‑inspired trend where unit costs fall while overall consumption rises. Goldman Sachs projects token usage to multiply 24‑fold between 2026 and 2030, driven by always‑on enterprise agents rather than higher query volumes [3]. If token‑price reductions outpace efficiency gains, total AI spending could still climb despite cheaper per‑token rates.

*Exact Terra rates not disclosed; inferred as “half” of Sol’s pricing per source [3].

What to watch

• July 9 public launch – confirms whether the limited preview expands to broader access.
• U.S. regulatory updates – any change to the executive order or export controls could alter rollout speed.
• Competitor pricing moves – Anthropic or Meta may adjust rates in response to OpenAI’s cuts.

OpenAI’s price war underscores a strategic bet that lower token costs will attract developers and enterprises, even as government oversight tempers immediate market impact. The key question remains whether the reduced rates will drive sustainable adoption or simply shift spending to higher‑token‑intensive agentic workloads.