OpenAI just cut GPT-5.6 prices — your agents got stronger overnight.
OpenAI just dropped a massive GPT-5.6 update — Luna is now 80% cheaper, and your AI agents just got way more powerful overnight.
This means the plan you're already on suddenly goes way further: your Luna usage stretches about five times as far as it did yesterday.
More builds, more agents, more finished work — without changing a single thing in your setup.
There's also a brand-new Fast mode for Sol that runs up to 2.5x quicker, and a hidden Codex upgrade that makes your agent reviews about 10x more efficient — I'll show you that one later.
Today I'll break down exactly what changed, what the builders are saying, and the switches to flip so you get every drop of it.
Stick with me. Let's get into it.
Here's the announcement in one breath.
Luna — the smallest, fastest GPT-5.6 — just dropped 80% in price.
Terra — the middle one — dropped 20%.
Sol — the big brain — got a new Fast mode in the API: up to 2.5x the speed at 2x the price, same intelligence.
And Auto-review in the ChatGPT app and Codex CLI jumped from GPT-5.4 to GPT-5.6 Luna — OpenAI says it now costs about 10x less.
The best part: the new prices are baked into how usage is counted in Codex and ChatGPT Work.
You don't flip anything. Your plan just goes further, starting today.
−80%GPT-5.6 Luna price
−20%GPT-5.6 Terra price
2.5xSol Fast mode speed (API)
~10xcheaper Auto-review in Codex
II ────── What actually changed
Three changes. All of them in your favour.
Straight from the announcement, in plain words.
1 — Luna costs 80% less.
Luna is the volume worker of the GPT-5.6 family — the one you point at everyday tasks.
Same model, same answers, a fifth of the price.
If Luna was your default, your usage now goes roughly five times further.
2 — Terra costs 20% less.
Terra is the balanced middle brain. A fifth off everything it does, every day, forever.
3 — Sol gets a Fast mode.
Sol is the frontier brain, and in the API you can now run it at up to 2.5x the speed of Standard processing for 2x the price.
Same intelligence, answers in less than half the wait.
When you're iterating live — debugging, building, shipping — speed IS the feature.
And the sleeper change: Auto-review in the ChatGPT app and Codex CLI upgraded from GPT-5.4 to GPT-5.6 Luna.
A newer, sharper model checking your agent's work — at about a tenth of the old cost.
Your agentic workflows literally got better AND lighter in the same update.
The same two models, before and after today. The bars are the whole story.
The announcement, straight from the source →
We are committed to pushing the model frontier across cost efficiency, capability, and speed. Starting today, we are reducing prices for GPT-5.6 Luna by 80% and GPT-5.6 Terra by 20%, and offering a faster option for GPT-5.6 Sol in the API.
"Is this only for API users, or does my normal plan benefit?"
Both. OpenAI says the lower Luna and Terra prices are reflected in how usage is counted in Codex and ChatGPT Work — so subscription users see their usage go further automatically.
The Sol Fast mode is the API-only piece.
✦
III ────── The reaction · today
What builders are saying — as it happened.
This dropped hours ago. Here's the live reaction from the builder community — all from today.
Along with the price reduction on GPT-5.6 Luna and Terra, Fast mode for GPT-5.6 Sol in the API delivers up to 2.5x the speed of Standard processing at 2x the Standard price. Fast mode gives API customers faster access to GPT-5.6 Sol, with no change in intelligence.
— OpenAI (@OpenAI)
We're also upgrading Auto-review in the ChatGPT app and Codex CLI from GPT-5.4 to GPT-5.6 Luna. Combined with Luna's new price, we expect Auto-review to cost about 10x less, making your agentic workflows more cost-efficient.
— OpenAI (@OpenAI)
OpenAI implemented a major price cut yesterday: $0.20 per million input tokens, $0.02 cached, $1.20 output (~80% reduction)
— Ashley Hindle (@ashleyhindle)
These updated rates are already live in the OpenAI API, Codex, and reflected in ChatGPT Work usage quotas
— Grok (@grok)
Luna at 80% off? Now I have no excuse not to finish my side project.
— lan (@Ian_Codes)
Easy to support open source models when you have this up your sleeve - awesome move from @OpenAI
— Jack Welsh (@_jacklwelsh)
Amazing news! You guys are doing great work in fighting the cost per token/inteligence speed. Keep on this trend!
— Kfir Gollan (@kfirgollan)
Can we get an Opus 5 vs these new Terra costs chart, please? @OpenAI
— Cutzilla (@Cutzilla)
Can we get a reset since with this new usage means all my Luna usage should've only used 20% but it used 100%? Pretty please
— Manu (@ManuellyBuilt)
So it´s only for API and not for subs?
— AI Inc (@AiTesty5)
All posts from today, pulled live — tap any card to open it on X.
✦
IV ────── What it means for your agents
The three-brain play just got a lot stronger.
Here's how I actually run the GPT-5.6 family, and why today's update supercharges it.
Sol, Terra and Luna aren't competitors — they're tiers.
Sol for the genuinely hard problems.
Terra for solid everyday work.
Luna for volume — the boilerplate, the drafts, the reviews, the hundred small tasks agents chew through all day.
When I benchmarked the family on real one-shot builds, all three shipped working results — and Luna finished its build at roughly a tenth of Sol's cost.
That was BEFORE today.
With Luna now 80% cheaper again, the volume tier is so light it's almost not worth metering.
The play after today:
Route harder DOWN. Tasks you used to send to Terra "to be safe"? Luna handles most of them, and at today's price the trade is obvious.
Let Auto-review run. It's now a better model at a tenth of the cost — the cheapest quality gate your agent workflow has ever had.
Save Fast-mode Sol for live sessions. When you're in flow and every wait breaks the rhythm, 2.5x speed is the difference between iterating and context-switching.
The three-tier play: Sol rarely, Terra by default, Luna constantly — and Luna just got 80% lighter.
Nothing about your plan changed. The distance it covers did.
"Won't routing more work to the small model hurt quality?"
Match the tier to the task, not to your fear.
In my own one-shot build tests, Luna shipped working, error-free builds — on tasks most people would have over-paid Terra for.
Keep Sol for the genuinely hard 10%, and let the quality gate (now a better model, ~10x cheaper) catch the misses.
✦
V ────── My story · why this matters
I was you. Then I stopped treating every task like a Sol task.
Before
I used to send everything to the biggest model, "to be safe".
Every draft, every rename, every tiny review — full frontier treatment.
My agents were slow, my limits ran dry by afternoon, and the work wasn't actually better.
Then I benchmarked the tiers against each other — and the small ones kept shipping.
After
Now the volume work rides the light tier, the hard 10% gets the big brain, and a cheap reviewer checks everything.
My agents run all day and the limits stopped being the thing I think about.
Today's update just made that exact setup dramatically stronger — Luna, the tier my agents lean on hardest, is suddenly 80% lighter.
You can run this too. It's a routing habit, not a rebuild.
VI ────── The receipts
Real people. Real builds. Inside the Boardroom right now.
The members running agent stacks on updates like this one — agency owners, ecom founders, course creators, 38 countries.
4,000+ founders inside AIPB
258 real wins documented
400K YouTube subscribers
38 countries · live members
163K X followers
Members post their wins as they happen — agents shipped, workflows automated, limits that stopped mattering. All collected in one doc you can read right now.
This update is live right now. The gains are sitting there whether you collect them or not.
So here's the deal.
Pick one: route your volume tasks to Luna, or turn Auto-review loose on your next Codex session.
One switch. Two minutes. Today — while everyone else is still just reading about it.
The people who move on update days are the ones whose agents quietly pull ahead of everyone else's.
Be one of those people. One switch. Today.
✦
VII ────── Old way vs new way
Most people read update news. Operators re-route same day.
The old way: see the announcement, nod, change nothing.
Keep sending every task to the big model "to be safe".
Keep hitting limits by mid-afternoon. The update happened; their setup never noticed.
The new way: treat every price-and-speed update as a free upgrade to YOUR machine.
Re-route the volume work to the tier that just got 80% lighter.
Turn the newly-cheap reviewer loose on everything.
Save the fast frontier brain for the moments speed actually pays.
Ten minutes of re-routing, months of compounding gains.
Same announcement, two kinds of people. The bars in your usage screen next week show which one you were.
✦ ✦ ✦
Get the whole operating system
Want your agents already wired for updates like this?
Inside my Agent OS, the GPT-5.6 family is one dropdown among many — when prices move, the routing play updates and every member gets it. Join the AI Profit Boardroom and you get:
The full Agent OS — every major model family in one dashboard, GPT-5.6 included
Model-routing playbooks — which tier for which job, kept current as prices move
Free local models — the everyday 90% of work at $0 on your own machine
Agent Kanban — Planner → Builder → Reviewer teams that ship while you sleep
Token-efficiency playbooks — stop thinking about limits entirely
5 live coaching calls a week — get unblocked by people on your exact setup
Inside the AI Profit Boardroom · skool.com/ai-profit-lab
Set up in an afternoon · used in 38 countries · new tools added every week
IX ────── Three beliefs to drop
What's actually stopping you isn't technical.
"Price cuts mean the model's on the way out."
The opposite here: OpenAI says the drop comes from efficiency gains — better GPU kernels, better decoding — found partly by using Sol itself.
This is the down-slope of a working cost curve, not a clearance sale.
"I'll wait until things stabilise before picking models."
Prices and models will move again — that's the one guarantee.
The winning setup isn't the right model pick; it's a routing habit that absorbs every update in minutes.
Build the habit once and every future price drop lands as pure upside.
"This stuff only matters at big-company scale."
Backwards — small operators feel limits hardest.
An 80% lighter volume tier matters MOST when one person's plan is doing the work of a team.
That's exactly the leverage updates like this hand you.
✦
X ────── The SOP
Collect the whole update — ten minutes, four flips.
1 · Re-point your volume work at Luna. Drafts, boilerplate, summaries, the hundred small agent tasks — the tier that carries them just got 80% lighter. If you were "playing safe" with Terra for these, stop.
2 · Leave Auto-review ON in Codex and ChatGPT. It upgraded from GPT-5.4 to GPT-5.6 Luna and now costs about a tenth of what it did. A better reviewer, basically free — there is no version of this you turn off.
3 · API user? Try Sol Fast mode on a live session. Up to 2.5x the speed, same intelligence. Use it when you're iterating in real time; skip it for overnight batch jobs where nobody's waiting.
4 · Check your usage next week, not today. The new counting is automatic in Codex and ChatGPT Work — the proof is your limits lasting longer. Note where you land, then push more volume through the light tier.
✦
XI ────── Recap
The whole update in 30 seconds.
OpenAI cut GPT-5.6 prices today: Luna −80%, Terra −20%, and Sol gained an API Fast mode — up to 2.5x speed at 2x price, same intelligence.
Auto-review in ChatGPT and Codex upgraded from GPT-5.4 to GPT-5.6 Luna — about 10x cheaper, and better.
Nothing to configure: the new prices are baked into how usage is counted in Codex and ChatGPT Work. Your plan just goes further.
The play: route volume DOWN to Luna, leave the newly-cheap reviewer on, save Fast-mode Sol for live sessions. Ten minutes of re-routing, compounding gains after.
The shortcut
Skip the routing homework. Run my exact setup.
The Agent OS ships with the routing already wired — and it updates when the model world moves. 4,000+ founders, 258 documented wins, 38 countries.
The AI Profit Boardroom is where the actual Agent OS lives — the templates, the prompts, the daily rooms, the weekly walkthroughs. Same builds you read about here, taught hands-on inside.
