GPT-5.6 Luna Just Got 80% Cheaper — Here's Why
Three weeks after launching GPT-5.6, OpenAI quietly rewrote its cheapest model's price tag. Luna dropped from $1 to 20 cents per million input tokens — an 80% cut that arrived faster and deeper than most price changes in this industry ever do. Here's what actually happened, why OpenAI moved this fast, and which workloads the new price genuinely opens up.
Luna's new rate makes high-volume, low-complexity workloads dramatically cheaper to run.
Luna Now Costs a Fraction of Its Launch Price
GPT-5.6 Luna, OpenAI's fastest and cheapest model in the GPT-5.6 family, dropped from $1.00 to $0.20 per million input tokens and from $6.00 to $1.20 per million output tokens on July 30, 2026 — an 80% cut across the board. GPT-5.6 Terra, the mid-tier model, fell 20%, from $2.50/$15 to $2/$12. The flagship model, Sol, kept its price but gained a new "Fast" mode that runs up to 2.5x quicker at twice the standard rate.
The change took effect immediately and applies automatically — no code changes, no opt-in. It also flows through to how usage is counted inside ChatGPT Work and Codex, so a fixed monthly credit budget now stretches noticeably further on tasks routed to Luna or Terra.
What Each GPT-5.6 Model Costs Now
GPT-5.6 Luna
$0.20 input / $1.20 output per million tokens, down from $1.00/$6.00. The fastest, cheapest model in the family — built for high-volume, lower-complexity work.
GPT-5.6 Terra
$2.00 input / $12.00 output per million tokens, down from $2.50/$15.00. OpenAI's "balanced" mid-tier model for everyday work.
GPT-5.6 Sol (Standard)
Unchanged at $5.00 input / $30.00 output per million tokens. Still the flagship for tasks that need the most capability.
GPT-5.6 Sol (Fast mode)
A new option, replacing the old Priority Processing tier: up to 2.5x faster than standard Sol, at double the standard price.
The Price War Behind the Timing
OpenAI's official framing pointed to efficiency gains — the company said improvements to its inference stack, including changes GPT-5.6 Sol made to its own production serving code, let it lower prices without cutting into the model's capability. But the timing lines up with a busier month than that explanation alone suggests.
July 24 — Anthropic ships Claude Opus 5
Anthropic's new flagship lands at $5/$25 per million tokens, delivering performance close to its top-tier Fable 5 model for roughly the same cost as its own predecessor.
Mid-July — Google ships Gemini 3.6 Flash
Google's workhorse-tier model arrives with an efficiency improvement of its own, adding pressure at the exact price point Luna and Terra compete in.
July 30 — OpenAI cuts Luna and Terra
One week after Opus 5, OpenAI moves on price rather than raw capability, undercutting the field at the low end of the market instead of trying to out-benchmark it.
Reporting from CNBC and others framed the move as a response to enterprise customers growing more cost-sensitive, particularly against competition from Chinese labs like DeepSeek shipping cheap, capable open models in the same window. Whatever the exact mix of motivations, the practical effect is the same: three major labs moved on price within about six weeks of each other.
Who Should Actually Switch to Luna
An 80% price cut is a big number, but it doesn't mean Luna is now the right model for everything. Luna sits at the bottom of the GPT-5.6 capability tier for a reason — OpenAI describes it as delivering roughly 85% of Sol's quality, not matching it.
For anything closer to the edge of what a model can reliably handle — complex multi-step reasoning, high-stakes accuracy requirements, nuanced writing — the price difference between Luna and Terra or Sol is still small relative to the cost of a wrong answer. Test your actual workload against your own evaluation criteria before moving production traffic down a tier just because the rate card changed.
Bottom Line
OpenAI cut GPT-5.6 Luna's price by 80% and Terra's by 20% on July 30, 2026, three weeks after launching the GPT-5.6 family — a fast, deep cut that landed in the middle of a broader round of price moves from Anthropic and Google. The change is automatic and applies to every API user immediately. It's a genuine win for high-volume, lower-complexity workloads, but it's not a reason to move accuracy-sensitive work down a model tier without testing first.
More AI Guides, Written Clearly
Browse our full library of AI tool guides, updated regularly as models and pricing change.
Common Questions
How much cheaper is GPT-5.6 Luna now?
GPT-5.6 Luna dropped 80% on July 30, 2026, from $1.00 to $0.20 per million input tokens and from $6.00 to $1.20 per million output tokens. GPT-5.6 Terra, the mid-tier model, fell 20%, from $2.50/$15 to $2/$12. The flagship, Sol, was not discounted.
Why did OpenAI cut Luna's price so soon after launch?
OpenAI said the cut reflects efficiency gains across its inference stack, including changes GPT-5.6 Sol made to its own serving code. The timing also followed price moves from Anthropic and Google within the same month, in what several outlets described as an intensifying price war among frontier labs.
Does the price cut affect my ChatGPT subscription?
No. ChatGPT Plus, Pro, Business, and Enterprise subscription prices are unchanged. The cut applies to API usage and to how Luna and Terra usage is counted against credit-based plans like ChatGPT Work and Codex, so a fixed budget now covers more output from those two models.
Is Luna still worth using over Terra or Sol?
For high-volume, lower-complexity tasks — classification, extraction, drafting, routine summarization — Luna's price makes it hard to beat on cost per task. For work that needs deeper reasoning or higher accuracy on hard problems, Terra or Sol still make more sense despite the higher rate.
More from the creator
Explore two other projects from the same creator.