OpenAI Cuts API Prices Again: What Changes for Small Builds
Cheaper tokens do not automatically mean a cheaper invoice. Two line items move your bill more than this cut does.
On this page
Another round of per-token price reductions has landed. Here's what it changes in practice for a small build.
#What actually moved
The reduction is asymmetric: input cheaper than output. That matters more than the headline number, because it decides who benefits.
- Benefits most: large prompt, short answer — classification, extraction, routing, moderation.
- Benefits least: short prompt, long answer — drafting, summarising into long form, code generation.
Check which shape your biggest job is before assuming the cut applies to you.
#Why your invoice may not fall
3 things routinely cancel out a price cut:
- Usage grows to fill the budget. Cheaper calls make it tempting to add a model step where a regular expression would do.
- Agent loops multiply everything. A per-call reduction applied to 6 times as many calls isn't a saving.
- Caching discounts are proportional, so if you were already caching, your absolute saving is smaller than the headline suggests.
#What to do this week
Pull last month's usage broken down by feature. If one background job is more than a third of the bill — and for most small teams, one is — look at whether it generates long outputs. If it does, this cut barely touches it, and batching will do far more.
#What this signals
Per-token pricing keeps falling and will keep falling. Build on the assumption that the model is the cheap part and your own orchestration — retries, loops, re-sent history — is where the money goes. That has been true for 2 years and nothing here changes it.
#What you should actually check this week
Pull 30 days of usage and sort it by feature before you do anything else. In most small teams 1 job is 40% or more of the bill, and it typically isn't the one you'd guess — it's a scheduled task nobody has looked at since the day it shipped. If that job generates long outputs, this price cut barely touches it and batching will do far more for you. The 20 minutes you spend reading your own usage breakdown is worth more than the cut itself.
Related reading: 6 ways to cut your bill without changing models.
Frequently asked questions
Should I switch models because of this?
No. A price cut is a reason to re-check your usage mix, not to change which model you call.
Does this affect the Batch API discount?
Batch pricing is derived from standard rates, so it moves too, but the relative discount stays roughly the same.
Written by
Vemuri Abhiram
Founder & Editor
I write every post here. The tools I review are ones I pay for and run on real client work, not demo accounts, and every number you see came off my own invoice or dashboard.
You might also like
EU AI Act Transparency Rules: What Small Teams Actually Have to Do
The EU AI Act transparency obligations that actually apply to small teams shipping AI features, in plain terms.
Vemuri Abhiram3 min read
How to Fix “Rate limit reached for requests” on the OpenAI API (6 Fixes)
The OpenAI rate limit error, how to tell which of the three limits you hit, and six fixes in the order that resolves it fastest.
Vemuri Abhiram6 min read
How to Fix “This model's maximum context length is…” (4 Fixes)
The context length error explained, plus four fixes ranked by how much answer quality each one costs — and why a bigger window should be last.
Vemuri Abhiram5 min read
More from Underread, once a week.
The guides worth reading and the tests worth trusting. One email, every Friday.
No spam. Unsubscribe in one click.