Price war: OpenAI cuts GPT-5.6 Luna by 80 percent
$0.20 instead of $1 per million input tokens: OpenAI slashes Luna prices – shifting competition from benchmarks to cost.
Illustration · AI-generated (AI IN LIFE)
At a glance
- GPT-5.6 Luna: from $1/$6 to $0.20/$1.20 per M input/output tokens (−80%)
- GPT-5.6 Terra: from $2.50/$15 to $2/$12 per M tokens (−20%)
- GPT-5.6 Sol: price unchanged, new Fast Mode at 2.5x speed
- OpenAI reports over 1 billion active users, ~50M paying subscribers
- 2025: about $13B revenue with continued heavy losses
OpenAI has cut API prices for GPT-5.6 Luna by 80 percent: a million input tokens now costs $0.20 instead of $1, while output tokens drop from $6 to $1.20. The larger GPT-5.6 Terra also got 20 percent cheaper – from $2.50/$15 to $2/$12 per million tokens. The flagship Sol keeps its price but gains a “Fast Mode” offering 2.5x speed at double the cost.
CFO Sarah Friar attributes the cuts to efficiency gains: “Better intelligence drives broader adoption. Broader adoption supports more investment.” At the same time, OpenAI announced it had passed one billion active users – with roughly 50 million paying subscribers at last count.
The context is an industry-wide price war: Chinese providers like DeepSeek won market share with aggressive pricing before DeepSeek itself recently hiked prices sharply. Observers see competition shifting – away from pure benchmark records, toward cost per completed task.
For developers and businesses, the message is clear: capable models are becoming a commodity, and prices for standard workloads keep falling. What remains open is how long providers can afford this subsidy logic – OpenAI still posted heavy losses in 2025 on about $13 billion in revenue.
FAQ
How much cheaper did GPT-5.6 Luna get?
80 percent: down to $0.20 per million input tokens and $1.20 per million output tokens.
Why is OpenAI cutting prices?
OpenAI points to efficiency gains, while also responding to price pressure from Chinese providers like DeepSeek.
What does this mean for users?
Standard AI workloads get much cheaper – competition is shifting from benchmarks to cost per completed task.
