One API. Every model. prices.
Access OpenAI, Anthropic, Google, DeepSeek and 40+ models through a single OpenAI-compatible endpoint — at the cheapest prices on the market with routing, fallbacks and usage analytics built in.
Cheapest
vs list price
40+
models
99.99%
uptime SLA
Supported models
Every frontier lab, one contract
Claude Fable 5.1
Anthropic
Input
$10.00/M
Output
$50.00/M
Context
1M
Claude Opus 5
Anthropic
Input
$5.00/M
Output
$25.00/M
Context
1M
GPT 6 Astra
OpenAI
Input
$10.00/M
Output
$50.00/M
Context
1M
Unified API
Drop-in OpenAI compatibility
Change two lines — baseURL and key — and every model speaks the same Chat Completions dialect. Stream, tool-call and batch exactly like you already do.
- ✓Single key for 40+ models
- ✓Streaming + tools + JSON mode
- ✓Automatic retries & fallbacks
import OpenAI from "openai";
const client = new OpenAI({
baseURL: "https://api.newtoken.ai/v1",
apiKey: process.env.NEWTOKEN_KEY,
});
const res = await client.chat.completions.create({
model: "auto", // routed by cost, latency & quality
messages: [{ role: "user", content: "Summarize this release" }],
});Smart routing
Route by price, latency or quality. Set policies per app, fall back automatically when a provider degrades.
auto · cost < $2/M · p99 < 800ms
Usage analytics
Per-key, per-model spend and token breakdowns with daily, weekly and monthly rollups. Export anywhere.
1.2M requests · $842.10 · this week
Developer experience
Playground, request logs, evals and typed SDKs. Ship behind a stable API while labs change underneath.
⌘K palette · logs · traces
Cheapest OpenAI-compatible API
NewToken is a unified AI API gateway for GPT-6, Claude, Gemini, DeepSeek, GLM, Kimi, Qwen, Grok and 230+ more models — one endpoint, per-token billing, up to 80% below list price. It is a drop-in OpenAI API alternative: keep your SDK, change the baseURL, and every chat completions call works with streaming, tools and JSON mode.
One API for every model
Instead of separate OpenAI, Anthropic, Google and DeepSeek accounts, NewToken routes each request to the cheapest online upstream route automatically. Compare live GPT-6 API pricing, Claude Opus and Claude Fable 5.1 prices, Gemini 3, DeepSeek and GLM-5.3 costs per 1M tokens in the catalog — with the discount versus the official vendor list price shown on every model.
Built for developers
Create API keys with per-key spend limits, pin specific routes per model, stream responses, and track request logs, spend by model, latency and tokens per second in the dashboard. Try every model free in the playground before writing a line of code, with temperature, reasoning effort and max-token controls.
Frequently asked questions
- What is NewToken?
- NewToken is an AI API gateway that gives you access to 230+ large language models — including GPT-6 Astra, Claude Fable 5.1, Claude Opus 5, Gemini 3 Pro, DeepSeek, GLM-5.3, Kimi K3, Qwen and Grok — through a single OpenAI-compatible API endpoint, at up to 80% below official list pricing.
- How is NewToken cheaper than using OpenAI or Anthropic directly?
- NewToken routes every request to the cheapest available upstream route for the selected model in real time, and passes the savings on. Prices are shown live per model with the discount versus the vendor list price. You pay per token with no seats, subscriptions or platform fees.
- Is the API compatible with the OpenAI SDK?
- Yes. NewToken is a drop-in replacement for the OpenAI API: point the OpenAI SDK's baseURL at https://api.newtoken.ai/v1 and use your NewToken key. Streaming, function calling, tools, JSON mode and chat completions all work unchanged.
- Which AI models are available on NewToken?
- Over 230 models from OpenAI (GPT-6, GPT-5.6), Anthropic (Claude Fable 5.1, Claude Opus 5, Claude Sonnet 5), Google (Gemini 3 Pro/Flash), DeepSeek, Z.ai (GLM-5.3), Moonshot AI (Kimi K3), Alibaba (Qwen 3.8), xAI (Grok 4.6), Meta (Llama), Mistral, MiniMax, NVIDIA and more. Selected models are free.
- How does pricing and billing work?
- You top up prepaid credits and are billed per 1M input and output tokens at the live rate shown on each model. Every key has configurable spend limits, and the dashboard shows per-request logs, spend by model, latency and token usage analytics.
- Can I try models before integrating?
- Yes — the built-in Playground lets you chat with any model in the catalog, tune temperature, max tokens and reasoning effort, attach files and see live streaming responses and cost per request, all billed like a normal API call.