LumeAPIInternal doc Q&A
45M input + 4M output tokens / month
25K tokens retrieved per query
- Official (Gemini 3.1 Pro)
- $138.00/mo
- LumeAPI
- $82.80/mo
- Monthly savings
- $55.2040% off
RAG API
RAG API — Document Q&A with Lower Input Token Costs
RAG resends retrieved chunks every query — lower input rates on Gemini Pro, Claude Sonnet, and GPT Terra cut document Q&A bills.
Model long-context capacity and input $/1M together when designing retrieval pipelines.
- OpenAI-compatible
- Pay as you go
- One API key
- GPT, Claude & Gemini
RAG-friendly model pricing
Official reference vs LumeAPI catalog rates. Pricing unit: per 1M input / output tokens. Last updated: July 2026. Source: provider list price.
Monthly cost examples
Illustrative totals for gemini-3.1-pro-preview using catalog list prices — your actual bill depends on retries, tool loops, and output length.
Customer knowledge base
120M input + 8M output tokens / month
High query volume RAG
- Official (Gemini 3.1 Pro)
- $336.00/mo
- LumeAPI
- $201.60/mo
- Monthly savings
- $134.4040% off
Retrieval dominates input
RAG API economics are usually input-heavy. Shrinking chunk size, caching embeddings, and summarizing history often beat switching models—but lower input list rates still matter at scale.
Compare Gemini Pro, Claude Sonnet, and GPT Terra on the same corpus with the same retrieval pipeline. Pick the model that meets quality bars at the lowest monthly input total.
Grounding features tied to Google Search may require native APIs; test compatibility early.
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
- Create an API key — register and open Console.
- Set the LumeAPI base URL — https://api.lumeapi.site/v1
- 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 (official RAG stack)
from openai import OpenAI
client = OpenAI(api_key="YOUR_OPENAI_API_KEY")
response = client.chat.completions.create(
model="gemini-3.1-pro-preview",
messages=[{"role": "user", "content": "Hello"}],
)After (LumeAPI)
from openai import OpenAI
client = OpenAI(
api_key="YOUR_LUMEAPI_KEY",
base_url="https://api.lumeapi.site/v1",
)
response = client.chat.completions.create(
model="gemini-3.1-pro-preview",
messages=[{"role": "user", "content": "Hello"}],
)Full step-by-step rollout, streaming checks, and FAQ: Prompt Caching guide — cut repeated input costs →
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
Caching retrieved context?
Application-side caching reduces repeat input tokens.
Model choice?
Balance context window, quality, and input $/1M.
Grounding?
Google Search Grounding may need native API.
Billing?
Input and output metered separately per call.
Built for these workloads
Document Q&A
Input-heavy — compare input $/1M across Pro, Sonnet, Terra.
Long context
Gemini Pro and Claude for large retrieved sets.
Retrieval token budget
Trim chunks and cache embeddings — API cost follows tokens sent.
Per-1000-query cost
Use scenario cards with your chunk size and query volume.
Related guides
FAQ
Best LLM API for RAG?
Often input-cost dominated — compare Gemini Pro and Sonnet input rates.
Cheap RAG API?
Lower LumeAPI input rates plus chunk optimization.
Long context pricing?
Linear in tokens sent — model docs for limits.
GPT vs Claude vs Gemini?
See table and run shadow queries on your corpus.
Gemini Pro RAG?
See /gemini-pro-api.
Agents with RAG?
See /ai-agent-api.
Reduce RAG input costs
Call retrieval-augmented prompts through https://api.lumeapi.site/v1. Need help choosing a model? Browse the developer docs or contact support.