Today we're launching NewToken — the cheapest way to call frontier AI models through a single API.
The problem
Every AI lab wants you to open an account, accept their terms, wire their SDK and top up their credits. OpenAI here, Anthropic there, Google somewhere else. If you want six models in production, you're maintaining six integrations, six billing relationships, and six sets of rate limits.
And the pricing you see on vendor pricing pages is rarely the pricing you can actually get — but you have no way to reach the better rates as a single developer.
What we built
One endpoint. NewToken is a drop-in OpenAI-compatible gateway. Point any OpenAI SDK at our base URL, swap the key, done:
python
from openai import OpenAI
client = OpenAI(
base_url="https://api.newtoken.ai/v1",
api_key="your-newtoken-key",
)
r = client.chat.completions.create(
model="claude-fable-5.1",
messages=[{"role": "user", "content": "Hello!"}],
)code
Streaming, tool calling, JSON mode — all of it works unchanged.
**230+ models.** GPT-6 Astra, Claude Fable 5.1, Claude Opus 5, Gemini 3 Pro, DeepSeek, GLM-5.3, Kimi K3, Qwen, Grok, Llama, and a long tail beyond that. Selected models are completely free.
**Cheapest-route routing.** Every request automatically goes to the cheapest online route for the model you picked. You can see the live price of every route in the catalog, including the discount versus the vendor's list price — we're transparent about the savings because the savings are the product.
**Pay per token.** No seats, no subscriptions, no platform fee. Top up credits, watch them drain at the rate shown on the model card. Per-key spend limits mean a leaked key can't ruin your month.
**A dashboard that respects you.** Usage analytics, spend by model, latency, tokens per second, per-request logs, a playground with temperature and reasoning-effort controls, and per-model route pinning when you want to pick your upstream yourself.
## Honest disclosures
We're a routing platform, not a model host. The models are served by third-party providers; we forward requests and bill tokens. That means:
- Model availability follows the upstream providers. A model can disappear or degrade without warning — we show route status live in the dashboard.
- Outputs are model outputs. They can be wrong. You're responsible for validating what you ship.
- Our prices include a margin over our upstream costs. We show list price and our price side by side, everywhere, always.
The full details live in our [legal center](/law).
## What's next
- **More routes, more models** — the catalog grows weekly.
- **API keys with granular spend limits** — already live, improvements coming.
- **Referral program** — invite friends, both sides get credits.
Go browse the [model catalog](/dashboard/models) or try something in the [playground](/dashboard/playground) — a few models are free, so it costs nothing to poke around.
— The NewToken Team