OpenAI dropped new API pricing and the developer timeline is absolutely cooked right now.
No cap, people woke up, checked their dashboards, and started doing math on napkins.
The headline number is the GPT-5 family, sitting at $1.25 per million input tokens and $10 per million output tokens.
Sounds cute until you remember output tokens are the expensive ones, and your chatbot is basically an output-token faucet that never turns off.
Meanwhile the older models got slashed hard.
GPT-4o dropped to $2.50 input and $10 output, and the mini tier is sitting at pocket-change prices.
Basically OpenAI is doing the classic "new flagship tax, old model clearance sale" move and honestly?
Cheaper tokens do not mean a cheaper bill, because cheap prices make everybody build greedier stuff.
Your app suddenly has memory, agents, background jobs, three retries, and a summarizer summarizing the summarizer.
That's not a discount, that's a subscription to vibes.
The real winners are indie devs and solo builders.
A one-person shop can now run a real product without a VC breathing down their neck.
The losers are the folks who hardcoded a budget spreadsheet in 2024 and never looked at it again.
Also, caching is the actual cheat code here.
If you're not caching your system prompts, you are lighting money on fire every single request and calling it innovation.
Same with batching, same with picking the dumbest model that still gets the job done.
And yeah, the competition is the reason this is happening.
Google, Anthropic, and a whole army of open-source models are all fighting for the same dev wallets, and price cuts are the loudest way to say "please don't leave." So what do you actually do?
Go pull your usage logs, find your top three most expensive endpoints, and ask if they truly need the flagship.
Half the time a mini model does the same job and you just never tested it because the big one felt safer.
Our take: this is genuinely good news, but only if you actually go look at your numbers.
Cheap tokens reward lazy architecture, and lazy architecture is how you end up paying more at a lower rate.
Final Thoughts
Go audit your stack before your next invoice does it for you.