OpenAI’s GPT-5.6 price cut landed on July 30, and the numbers are honestly kind of wild.
Luna dropped to $0.20 per million input tokens. That’s an 80% slash. Terra got trimmed by 20%. And Sol? Still sitting at $5/$30. But now there’s a Fast mode that runs 2.5× faster for double the price. (F2, F3, F4, F5)
Why the Cut Hits Different for Solo Devs
The cheapest tier just became disposable. At $0.20 per million input tokens, a 10k-token request costs $0.002. Pennies. You could run that hundreds of times daily and still not crack a modest budget.
My agency uses Luna for nightly compliance reports. Cost per run fell from roughly $0.10 to $0.02. That’s not a rounding error. It’s real budget back. We’ve already redirected it into extra monitoring jobs. (F3, F8)
What Fast Mode Means for Agent Loops
Sol’s Fast mode promises 2.5× speed at double the $5/$30 rates. OpenAI says the underlying intelligence is unchanged. You’re not paying more for smarts, you’re paying more for urgency. That distinction matters if you’re building agents that loop, retry, or chain calls.
Honestly, Fast mode feels aimed at production systems where latency is the bottleneck, not side projects.
For agent workflows that fire dozens of calls in sequence, a 2.5× speedup can shave meaningful time off end-to-end execution. But it’s not free. Double the price stacks fast if your loop is chatty.
I’d probably only flip it on for the critical path, not every tool call.
side note: their docs are still a mess for toggling model variants programmatically.
Price Changes at a Glance
Here’s what actually changed starting July 30:
– Luna input tokens: cut to $0.20 per million. An 80% drop
– Terra: 20% cheaper across the board
– Sol: still $5/$30, but Fast mode now runs 2.5× faster for double the price
Use this if you’re doing quick math on a project budget or trying to justify the switch to a skeptical cofounder.
FAQ: GPT-5.6 API Price Update
Did all GPT-5.6 models get cheaper?
No. Luna and Terra got cheaper. Sol didn’t drop in price. It gained a Fast mode option instead.
When did the new pricing take effect?
July 30.
How much cheaper is Luna now?
Luna is $0.20 per million input tokens, which is an 80% reduction.
What does Sol Fast mode cost?
Double the standard Sol rate of $5/$30.
Is Sol Fast mode actually faster?
OpenAI says it runs 2.5× faster, with the same underlying intelligence.
Should I switch everything to Luna?
If your workload can tolerate the cheapest tier, probably. But don’t force it if you need higher capability — Luna is cheap for a reason.
Wrapping Up
The GPT-5.6 update is mostly a win for builders watching their token bill.
Luna at $0.20 per million tokens changes the math for low-stakes, high-volume work. Terra’s 20% cut helps mid-tier loads. Sol’s Fast mode gives you a latency lever when you need it. Just remember it costs double. My take? Start with Luna, benchmark your actual usage, and only pay up for Fast mode when a delay actually hurts.
