LumeAPI
Get started

Multi-model API

Multi-Model AI API — One Key for GPT, Claude & Gemini

One LumeAPI account, one API key, one wallet, and one request format for GPT, Claude, Gemini, image, and video models.

Unlike /openai-compatible-api (SDK migration), this page focuses on unified operations — switch models by changing the model id only.

  • OpenAI-compatible
  • Pay as you go
  • One API key
  • GPT, Claude & Gemini
Create API keyView pricingRead docs

Flagship models on one gateway

Official reference vs LumeAPI catalog rates. Pricing unit: per 1M input / output tokens. Last updated: July 2026. Source: provider list price.

ModelOfficial (in / out)LumeAPI (in / out)Savings
GPT-5.6 Terragpt-5.6-terra$2.50 / $15.00$1.25 / $7.5050% offDetails →
Claude Sonnet 4.6claude-sonnet-4-6$3.00 / $15.00$1.80 / $9.0040% offDetails →
Gemini 3.1 Progemini-3.1-pro-preview$2.00 / $12.00$1.20 / $7.2040% offDetails →
Gemini 3.5 Flashgemini-3.5-flash$1.50 / $9.00$0.90 / $5.4040% offDetails →

Monthly cost examples

Illustrative totals for gpt-5.6-terra using catalog list prices — your actual bill depends on retries, tool loops, and output length.

Product with model routing

30M input + 10M output tokens / month

GPT, Claude, and Gemini on one client

Official (GPT-5.6 Terra)
$225.00/mo
LumeAPI
$112.50/mo
Monthly savings
$112.5050% off

Rates last updated July 2026

Agent orchestration

90M input + 30M output tokens / month

Different models per agent step

Official (GPT-5.6 Terra)
$675.00/mo
LumeAPI
$337.50/mo
Monthly savings
$337.5050% off

Rates last updated July 2026

Operations, not just SDK syntax

Multi-model API value is operational: one account, one wallet, one Usage export, and one integration path for GPT, Claude, Gemini, image, and video. Engineering changes model id per feature flag; finance sees a single ledger.

This differs from /openai-compatible-api, which answers “will my OpenAI client work?” Multi-model answers “can I run my product without three vendor relationships?”

Route free-tier users to Flash or mini models and paid tiers to Sonnet or Terra without forking your codebase—only configuration changes.

Self-serve path: register to first API call

LumeAPI is designed for developers who want to integrate without scheduling demos. Create an account, confirm your email, and open Console to generate an API key. Fund your USD wallet with USDT on supported chains when you are ready for billable traffic—there is no mandatory minimum beyond what your tests require.

Point your OpenAI-compatible client at https://api.lumeapi.site/v1, set Authorization to Bearer your key, and pass a catalog model id in the model field. Run a short curl or SDK script from /docs to verify latency, streaming, and error handling before you attach the key to production services.

Use Usage logs to reconcile per-call cost with finance forecasts. When a model tier is too expensive or quality is insufficient, change model id—not your entire integration. For cross-provider price tables and Research deep dives, follow internal links on this page rather than duplicating migration math here.

Documentation, catalog, and support

Every catalog model has a detail page under /models with official reference pricing, LumeAPI pricing, and links to /docs/models/{id} for parameters and curl examples. Start there when this commercial page points you to a model id you have not called before.

The /docs index lists gateway authentication, Chat Completions, image endpoints, and async video patterns. llms.txt bundles the same information for agent tooling—useful when you want a single URL to paste into Cursor or an internal bot.

Research articles explain why bills grow and how to compare providers; commercial pages like this one explain what LumeAPI offers and how to start. Follow internal links instead of searching for duplicate migration content across pages.

If billing, chain deposits, or integration behavior is unclear, use /contact for support channels. Include your model id, approximate request time, and whether the issue is authentication, balance, or model parameters—that speeds up resolution.

Why LumeAPI

Mainstream models only

GPT, Claude, and Gemini catalog ids with published per-token or per-media rates — not obscure small models marketed as discounts.

Transparent pricing

Official reference vs LumeAPI columns on every commercial page. Your request model id matches Usage logs and billing.

OpenAI-compatible gateway

One base URL for Chat Completions, image generations, and async video. Swap key, base URL, and model id — keep your SDK.

Self-serve wallet

Top up with USDT, create keys in Console, and track per-call cost in Usage — no sales calls required.

Get started in three steps

  1. Create an API key — register and open Console.
  2. Set the LumeAPI base URL — https://api.lumeapi.site/v1
  3. Choose a supported model id — from the table above or model catalog.

Migrate in minutes

Three values change: API key, base URL, model id. Everything else stays the same.

Before (three provider integrations)

python
# Separate SDKs per provider — three keys and billing relationships

After (LumeAPI)

python
from openai import OpenAI

client = OpenAI(api_key="YOUR_LUMEAPI_KEY", base_url="https://api.lumeapi.site/v1")

for model in ("gpt-5.6-terra", "claude-sonnet-4-6", "gemini-3.5-flash"):
    r = client.chat.completions.create(
        model=model,
        messages=[{"role": "user", "content": "Reply with your model name only."}],
    )
    print(model, r.choices[0].message.content)

Full step-by-step rollout, streaming checks, and FAQ: Model routing guide — quality, speed and reliability →

Migration & compatibility

What changes

API key, base URL (https://api.lumeapi.site/v1), and model id to a LumeAPI catalog entry. Message shape stays OpenAI-compatible for most apps.

What to test

Streaming, tool calling, JSON mode, and error handling on your heaviest models. Shadow 5–10% of traffic before full cutover.

What may differ

Provider-native features (Anthropic Batch, Google Grounding, OpenAI Assistants) may require the official API. Test your exact payload.

Rollback

Keep environment variables for base URL and model id. Switch back instantly if staging tests fail.

Trust & billing

One key for all models?

Yes. Keys call any catalog model your balance supports.

Separate wallets per model?

No. One USD wallet. Usage logs show model id per request.

How to switch models?

Change the model field in Chat Completions. No new endpoint.

Limit models per key?

Use separate keys per environment for operational separation.

Built for these workloads

AI SaaS with tiered models

Free tier on Flash, paid on Sonnet — same integration code.

Agent platforms

Users pick GPT, Claude, or Gemini without three vendor accounts.

Internal tools

One Usage export. Swap models with a config change.

Multimodal apps

Text, image, and video on one key and wallet.

Related guides

FAQ

Different from OpenRouter?

LumeAPI focuses on mainstream models at lower catalog rates.

Different from /openai-compatible-api?

Compatible = SDK migration. Multi-model = unified account and wallet.

Need three provider keys?

No. One LumeAPI key for supported catalog models.

A/B test models?

Yes. Route by model id and compare Usage logs.

Image and video included?

Yes. Browse /models for media ids.

Compare prices?

See /ai-api-pricing.

Unify your model stack today

One key calls every row below through https://api.lumeapi.site/v1. Need help choosing a model? Browse the developer docs or contact support.

Create API keyOpen consoleModel catalog