If you're building AI wrappers, agents or side projects, your LLM API bill is probably your biggest cost. Here's how to cut it hard — without changing your code.
The problem with list pricing
OpenAI, Anthropic and Google all bill per token at list price. OpenRouter aggregates them, but its markup stacks on top. Meanwhile indie hackers ship products where a $0.40-per-request model quietly eats the entire margin.
Cheapest-route routing (the big win)
A model like Claude Opus 5 or GPT-6 is not one deployment — it's served through multiple upstream routes with different costs and availability. NewToken's gateway routes every request to the cheapest currently-online route for that model, automatically, and shows you the discount versus list price on every model card.
The result: up to 80% below list price on selected models, with the same API contract.
A discounted OpenAI API key that isn't sketchy
You don't need to buy gray-market keys. NewToken issues you a normal API key, you point the OpenAI SDK's baseURL at https://newtoken.dev/v1, and every chat completions call — streaming, tools, JSON mode — works unchanged. You just pay NewToken's rate instead of list.
from openai import OpenAI
client = OpenAI(
base_url="https://newtoken.dev/v1",
api_key="your-newtoken-key",
)## Free models for prototyping
Selected models on NewToken are completely free — enough to prototype an entire product before you pay a cent. The playground lets you run multiple LLMs side by side with temperature and reasoning-effort controls, so you can benchmark quality per dollar before committing.
## Spend limits save careers
Every key has a configurable spend limit. A leaked key drains to the limit and stops — your month is not ruined.
## When is NewToken NOT the right choice?
- You need a dedicated single-vendor SLA with the lab itself.
- You need fine-tuned private deployments (not offered).
- You need vendor-native features that only exist on the first-party API.
For everything else — wrappers, agents, chatbots, RAG pipelines, eval harnesses — a routing gateway with live price comparison is simply cheaper.
Browse the [model catalog](/dashboard/models) and check the live discount on the models you already use.