AI Lab · by Tiny Mango · 2026-10-17

The cost of inference is dropping, but my bill keeps growing

Per-token prices fell a lot, yet usage exploded because we added agents and retries. How do you control spend?

33 sparks · 19 comments

Comments

Bilal Raza · 14 likes

Managing inference costs sounds simple until you try it with real constraints.

I'm more of an enthusiast than a pro, but experimenting with inference costs on weekends has been fun.

Yui Tanaka · 0 likes

we saw the opposite honestly

My manager read one article on inference cost and now it's our whole roadmap lol

Minji Kim · 5 likes

I've lived this exact meeting

Would love to see actual data on inference cost instead of vibes. Anyone have links

yeah thats what I meant

Fair enough

cost vs benefit matters here, we ran numbers on inference cost and the payoff came later than expected!

agree mostly, though not for every case

Fair enough

Quiet Fern · 1 likes

Learning about inference costs is easier than a year ago. That part is underrated.

Good point, didn't think of that

we saw the opposite honestly

I do this for a living, and the biggest costs are the things nobody puts in the demo.

lol my cat has stronger opinions on it than I do

my wallet felt that

Emir Kaya · 0 likes

We tried inference cost last quarter, worked great for small stuff, got messy at scale

send help