Live · 103 models · 99.00% uptime

One key.
Every frontier model.

OpenAI, Anthropic, Google, Meta, Mistral, xAI — routed through a single OpenAI-compatible endpoint. Switch models without switching SDKs. Pay only for what you use.

Models routed103
Uptime, 30d99.00%
p50 latency412 ms

What you get

Built for builders who ship before the next release lands.

OpenAI-compatible

Drop-in for the openai SDK. Change base_url, keep your code. Streaming, tool calls, vision — all parity.

Compatible with openai>=1.0

Model freedom

GPT-5, Claude Opus 4.8, Gemini 2.5, Llama 4, Grok 4, Mistral Large, Kimi K2 — call any of them by name.

103 models, growing weekly

Predictable pricing

Pay-as-you-go credit. No subscriptions, no platform fee, no overage surprises. Credit never expires.

$10 minimum top-up

Built-in failover

If a provider degrades, we route to the next-best model in-family. Your request still completes.

Sub-second cutover

Per-key budgets

Hard cap on every key. Spin up keys per project, per developer, per CI job — no surprise bills.

Hard limits, not soft alerts

EU-hosted gateway

Routed through Sweden Central. Behind Cloudflare. TLS 1.3 end-to-end. No third-party logging hooks.

Azure Sweden + Cloudflare

Drop-in

One line changes.
Then every model.

If you have the openai SDK installed, you already have Meridian support. Swap base_url, keep the rest of your code.

  • Streaming, tools, vision, JSON mode — full parity.
  • Switch models by changing the string. No re-write.
  • Works with LangChain, LlamaIndex, Vercel AI SDK, Anthropic SDK.
meridian_quickstart.py
from openai import OpenAI

client = OpenAI(
    api_key="sk-nimbus-...",                       # your Meridian key
    base_url="https://llm.getnimbus.net/v1",       # one endpoint
)

# Use any model name from the catalog
response = client.chat.completions.create(
    model="claude-opus-4-8",                       # or gpt-5, gemini-2.5-pro, grok-4
    messages=[{"role": "user", "content": "ship it"}],
)

Pricing

Pay once. Use until empty.

Top up a key with the amount you want. Credit never expires. No subscription, no auto-renew.

Want something custom? Top up any amount with crypto.

Questions

The short version.

You get one key for every frontier model — OpenAI, Anthropic, Google, Meta, Mistral, xAI — instead of managing five vendor accounts, five billing cycles, five SDKs, and five sets of rate limits. Switch models with a string change. Add failover with a config flag.

Yes. Point any OpenAI SDK at https://llm.getnimbus.net/v1 with your Meridian key and it works. Streaming, function calling, vision, JSON mode, tool use — all supported.

Requests return a 429 with a clear message. Your key stays valid — top it up and it keeps working. No auto-charge, no surprise renewal.

We store request and response payloads for billing reconciliation, abuse detection, and platform reliability. We never sell, share, or train on customer data. Full policy is in our Privacy page.

103 routed at launch. The full live list is in Models, including pricing per million tokens for each. We add new releases within 24-48 hours of provider launch.

Yes. The gateway runs on Azure Sweden Central behind Cloudflare with active-active failover. Uptime is published live in the header. Per-key budgets prevent runaway spend.

Credit card via SellAuth, crypto via NOWPayments or Coinbase Commerce. No invoice required for any pack under $500.