We ran into the same issue building agent workflows, which is why we started building https://oxlo.ai — experimenting with a flat subscription model where we absorb the token variance so builders don’t have to constantly model token risk.

We ran into the same issue and ended up building https://oxlo.ai to make the cost side more predictable for agent workloads.

A single user action can trigger anywhere from a few to dozens of LLM calls (tool use, retries, reasoning steps), and with token-based pricing the cost can vary a lot.

How are builders here planning for this when pricing their SaaS?

Are you just padding margins, limiting usage, or building internal cost tracking? Also curious, would a service that offers predictable pricing for AI APIs (like a fixed subscription cost) actually be useful for people building agentic workflows?

Comments URL: https://news.ycombinator.com/item?id=47332177

Points: 5

# Comments: 23

I’m one of the founders of Oxlo.ai. We’re building a developer first AI API platform focused on simplifying how small teams integrate AI into production.

Most AI APIs charge per token, which can make costs unpredictable as usage grows. Oxlo.ai takes a different approach: request based pricing with unlimited token output per request.

We provide unified API access to curated open models across: • Text generation • Coding • Embeddings • Image generation • Audio & speech • Computer vision

The goal isn’t to replace large API providers. If you’re already using a major API in production, Oxlo.ai can act as a complementary layer.

For example, teams can: • Route simpler or lower-priority workloads to Oxlo.ai under predictable pricing • Keep higher-complexity or overflow workloads with their existing provider • Implement fallback routing when one endpoint is busy

This hybrid approach can improve cost control while maintaining production reliability.

We’re still early (<3k users) and actively looking for feedback, especially from teams running AI features in production.

Happy to answer questions.

— Barath Kanna - Founder, Oxlo.ai

Comments URL: https://news.ycombinator.com/item?id=47128001

Points: 1

# Comments: 1

While this isn’t hard to understand, in practice it makes costs harder to predict, especially for small teams moving from experiments to early production. This feels less like a technical problem and more like a budgeting and planning problem.

I’m curious about alternative pricing abstractions, for example a subscription with unlimited tokens but a capped number of requests, aimed at making monthly spend easier to reason about while building.

For people running AI in production today, does token based billing give you enough predictability, or would a model like this actually reduce friction? What tradeoffs would matter most to you?

Comments URL: https://news.ycombinator.com/item?id=46781011

Points: 1

# Comments: 4

These are order-of-magnitude numbers, but the takeaway is that multi H100 boxes are plausibly ~100× faster than workstation Macs for this class of model, especially for long-context prefill.