Gateway responding

Chat · API · CLI

AI codingwithoutlimits.

Claude Opus 5.5, Claude Fable 5.2 and Claude Sonnet 5. Unlimited use from $31.10 USD/mo. never per token.

Subject to our fair use policy — see Terms.
ClaudeGPTGLMHunyuanDeepSeekMiMo

VisaMastercardCard · instant activation

lumiarouter.comexample session
export ANTHROPIC_BASE_URL=https://lumiarouter.com
claude
claude-opus-5-5
Refactor the auth middleware
14 files read · 3 changed
/model claude-sonnet-5
same history, another model
Write the middleware tests
6 tests created · all passing
messages this month12,486
extra charges$0.00 USD
Card · instant activation
PythonNode.jsClaude Code · Codex

no surprise bills

  • 10models in one key
  • 1Mtokens of context
  • 4API formats
  • 98.94%uptime, last 30 days

Works with your tools

  • Claude Code
  • Codex CLI
  • Cursor
  • Cline
  • VS Code
  • Windsurf
  • OpenCode
  • Open WebUI
  • n8n
  • AnythingLLM
  • Zapier
  • Make

Frontier models,one API key.Frontier models,one API key.

Drag sideways to see every model →

Claude Opus 5.5AnthropicClaude Sonnet 5AnthropicClaude Sonnet 5.5AnthropicClaude Fable 5.2AnthropicGLM 5.3Z.aiGLM 5.3 FlashZ.aiHunyuan Hy4 PreviewTencentDeepSeek V4.1 FlashDeepSeekMiMo V2.6 ProXiaomiGPT-6 AstraOpenAI
Best for

Best balance of capability and cost

Read specifications

The workhorse of agentic coding: multi-file refactors, end-to-end features and long sessions without losing the thread.

Reasoning · Agentic · Code · Vision · Long context

Provider list price per 1M tokens: US$ 5 / US$ 25 (input / output).

The fastest without giving up quality

Read specifications

The balance of speed and capability: answers fast, follows instructions precisely and handles most of the day-to-day work without the wait.

Reasoning · Agentic · Code · Vision · Long context

Provider list price per 1M tokens: US$ 3 / US$ 15 (input / output).

Fast and more capable than Sonnet 5

Read specifications

The new generation of Sonnet: the same response speed, with more capability for code and everyday agentic work.

Reasoning · Agentic · Code · Vision · Long context

Provider list price per 1M tokens: US$ 3 / US$ 15 (input / output).

Maximum capability available

Read specifications

Anthropic's most capable model. Reasoning always on and stamina for hours of autonomous work. for when the problem is genuinely hard.

Reasoning · Agentic · Code · Vision · Long context

Provider list price per 1M tokens: US$ 10 / US$ 50 (input / output).

Long-running agentic work with a 1M context window

Read specifications

OpenAI’s frontier model for long-running agentic work, with a 1M context window.

Reasoning · Agentic · Code · Long context

Provider list price per 1M tokens: US$ 0.6 / US$ 2.2 (input / output).

Long-running agentic work with a 1M context window

Read specifications

OpenAI’s frontier model for long-running agentic work, with a 1M context window.

Reasoning · Agentic · Code

Provider list price per 1M tokens: US$ 0 / US$ 0 (input / output).

Long-running agentic work with a 1M context window

Read specifications

OpenAI’s frontier model for long-running agentic work, with a 1M context window.

Reasoning · Code

Provider list price per 1M tokens: US$ 0 / US$ 0 (input / output).

Long-running agentic work with a 1M context window

Read specifications

OpenAI’s frontier model for long-running agentic work, with a 1M context window.

Reasoning · Code

Provider list price per 1M tokens: US$ 0 / US$ 0 (input / output).

Long-running agentic work with a 1M context window

Read specifications

OpenAI’s frontier model for long-running agentic work, with a 1M context window.

Reasoning · Code

Provider list price per 1M tokens: US$ 0 / US$ 0 (input / output).

Long-running agentic work with a 1M context window

Read specifications

OpenAI’s frontier model for long-running agentic work, with a 1M context window.

Reasoning · Agentic · Code · Long context

Provider list price per 1M tokens: US$ 2.5 / US$ 12 (input / output).

Context256kStarter / Pro / Max1MPro and Max256kStarter / Pro / Max1MPro and Max256kStarter / Pro / Max1MPro and Max256kPro / Max1MPro and Max1MAll plans1MAll plans1MAll plans1MAll plans1MAll plans1MPro and Max
Output128K128K128K128K128K8K8K8K8K128K
Speed
Quality
Planincluded from Starterincluded from Starterincluded from Starterincluded from Proincluded from Maxincluded from Starterincluded from Starterincluded from Starterincluded from Starterincluded from Starter
  • Claude Opus 5.5

    Anthropic

    Best balance of capability and cost

    About this model

    The workhorse of agentic coding: multi-file refactors, end-to-end features and long sessions without losing the thread.

    Reasoning · Agentic · Code · Vision · Long context

    Provider list price per 1M tokens: US$ 5 / US$ 25 (input / output).

    Context
    256kStarter / Pro / Max1MPro and Max
    Output
    128K
    Speed
    Quality
    Plan
    included from Starter
  • Claude Sonnet 5

    Anthropic

    The fastest without giving up quality

    About this model

    The balance of speed and capability: answers fast, follows instructions precisely and handles most of the day-to-day work without the wait.

    Reasoning · Agentic · Code · Vision · Long context

    Provider list price per 1M tokens: US$ 3 / US$ 15 (input / output).

    Context
    256kStarter / Pro / Max1MPro and Max
    Output
    128K
    Speed
    Quality
    Plan
    included from Starter
  • Claude Sonnet 5.5

    Anthropic

    Fast and more capable than Sonnet 5

    About this model

    The new generation of Sonnet: the same response speed, with more capability for code and everyday agentic work.

    Reasoning · Agentic · Code · Vision · Long context

    Provider list price per 1M tokens: US$ 3 / US$ 15 (input / output).

    Context
    256kStarter / Pro / Max1MPro and Max
    Output
    128K
    Speed
    Quality
    Plan
    included from Starter
  • Claude Fable 5.2

    Anthropic

    Maximum capability available

    About this model

    Anthropic's most capable model. Reasoning always on and stamina for hours of autonomous work. for when the problem is genuinely hard.

    Reasoning · Agentic · Code · Vision · Long context

    Provider list price per 1M tokens: US$ 10 / US$ 50 (input / output).

    Context
    256kPro / Max1MPro and Max
    Output
    128K
    Speed
    Quality
    Plan
    included from Pro
  • GPT-6 Astra

    OpenAI

    Long-running agentic work with a 1M context window

    About this model

    OpenAI’s frontier model for long-running agentic work, with a 1M context window.

    Reasoning · Agentic · Code · Long context

    Provider list price per 1M tokens: US$ 2.5 / US$ 12 (input / output).

    Context
    1MPro and Max
    Output
    128K
    Speed
    Quality
    Plan
    included from Max
  • GLM 5.3

    Z.ai

    Long-running agentic work with a 1M context window

    About this model

    OpenAI’s frontier model for long-running agentic work, with a 1M context window.

    Reasoning · Agentic · Code · Long context

    Provider list price per 1M tokens: US$ 0.6 / US$ 2.2 (input / output).

    Context
    1MAll plans
    Output
    128K
    Speed
    Quality
    Plan
    included from Starter
  • GLM 5.3 Flash

    Z.ai

    Long-running agentic work with a 1M context window

    About this model

    OpenAI’s frontier model for long-running agentic work, with a 1M context window.

    Reasoning · Agentic · Code

    Provider list price per 1M tokens: US$ 0 / US$ 0 (input / output).

    Context
    1MAll plans
    Output
    8K
    Speed
    Quality
    Plan
    included from Starter
  • Hunyuan Hy4 Preview

    Tencent

    Long-running agentic work with a 1M context window

    About this model

    OpenAI’s frontier model for long-running agentic work, with a 1M context window.

    Reasoning · Code

    Provider list price per 1M tokens: US$ 0 / US$ 0 (input / output).

    Context
    1MAll plans
    Output
    8K
    Speed
    Quality
    Plan
    included from Starter
  • DeepSeek V4.1 Flash

    DeepSeek

    Long-running agentic work with a 1M context window

    About this model

    OpenAI’s frontier model for long-running agentic work, with a 1M context window.

    Reasoning · Code

    Provider list price per 1M tokens: US$ 0 / US$ 0 (input / output).

    Context
    1MAll plans
    Output
    8K
    Speed
    Quality
    Plan
    included from Starter
  • MiMo V2.6 Pro

    Xiaomi

    Long-running agentic work with a 1M context window

    About this model

    OpenAI’s frontier model for long-running agentic work, with a 1M context window.

    Reasoning · Code

    Provider list price per 1M tokens: US$ 0 / US$ 0 (input / output).

    Context
    1MAll plans
    Output
    8K
    Speed
    Quality
    Plan
    included from Starter

What would you spendpaying per token?What would you spendpaying per token?

Model
PAYING PER TOKEN

Claude Opus 5.5 · 4,000 messages

input3.6M tokens × US$ 5$18.00
output2.8M tokens × US$ 25$70.00
total$88.00
in USD$88.00 USD

LumiaRouter Pro(30 days)$31.10 USD
API access(30 days)$20 USD
Monthly difference$49.00 USD

900 input tokens and 700 output tokens per message.

Estimate based on provider list prices.

PAYING PER TOKEN: $88.00 USD. Monthly difference: $49.00 USD.

Get Pro
FEATURES

Built for people who use AI all day

No frills: the things that matter once AI becomes part of your job.

Only the top, no filler

Frontier models only, picked for code. No "budget" option for you to pick wrong halfway through the job.

Claude Opus 5.5Claude Sonnet 5Claude Sonnet 5.5Claude Fable 5.2GLM 5.3GLM 5.3 FlashHunyuan Hy4 PreviewDeepSeek V4.1 FlashMiMo V2.6 ProGPT-6 Astra

Genuinely unlimited

No message quota, no per-token billing, and none of that dread when the invoice lands.

Messages this month12,512
Extra charges$0.00 USD

A second opinion in one click

Stuck with one model? Switch to another lab mid-conversation, same history intact.

Streaming responses

Text arrives token by token, the same in the chat and over the API. First token in under 300ms.

1M context

Whole repository, full build log, a migration across dozens of files — all in one conversation.

repository · 410kbuild log · 180kdocs · 120k710k / 1M

Compatible API

The chat-format endpoint you already use. Switching from your current provider is a base URL and a key.

Your code stays yours

Nothing you send trains a model. History lives in your account and you can delete it any time.

Full history

Every conversation is saved, searchable, pinnable and organised. Go back to any thread whenever.

Search chats…Refactor auth middlewarenowFix flaky CI test3hPostgres migration plan6h

How it works

  1. Choose a plan

  2. Generate a key in your dashboard

    sk-•••••••••••• Get my key

    API access · $20 USD per 30 days

  3. Paste your key into Claude Code or Codex

    $env:ANTHROPIC_BASE_URL="https://lumiarouter.com"

Works where you already work

  • Claude Code
  • Codex CLI
  • Cursor
  • Cline
  • VS Code
  • Windsurf
  • OpenCode
  • Open WebUI
  • n8n
  • AnythingLLM
  • Zapier
  • Make
API

The same account in your own apps

A chat-format endpoint you already know, with SSE streaming. Generate a key in the dashboard, swap the base URL, and your code stays the same — now with every model in the catalog. API access is included in every plan.

  • Works with the SDKs you already use
  • Token-by-token streaming over SSE
  • One key, every model on your plan
  • Multiple keys per project on Max
lumiarouter.com/v1200 OK
curl /v1/chat/completions \
  -H "Authorization: Bearer $LUMIA_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-opus-5-5",
    "stream": true,
    "messages": [{"role": "user", "content": "Explica RAG em 3 linhas"}]
  }'
  • Anthropic/v1/messages
  • OpenAI/v1/chat/completions
  • SSEstream: true

Why is it so much cheaper?Why is itso much cheaper?

LumiaRouter is a hub of AI models. We buy in volume from providers, make use of credits and quotas, and route you to the best path to reach the models. It is the same Claude. it just fits your budget.

What about quality?

It is Claude itself, unchanged. We do not alter the model. What can vary at times is response speed at peak hours. and even then it is fast.

Works everywhere

Use it in Claude Desktop (Windows, Mac, Linux), Claude Code (terminal), the browser, or any compatible tool. The same key works in all of them.

Fixed price. Unlimited use.Fixed price.Unlimited use.

VisaMastercard30-day plans · card or crypto

No prepaid credits, no per-token billing, no small print. Cancel whenever you want, right from the dashboard.

Pro

Ideal for people who use AI every day for work and study.

US$31.10/mo

  • Use as much as you want, no limits
  • Claude Opus 5.5 and Claude Sonnet 5
  • Switch between models anytime
  • Access on your computer and browser
  • Fast responses, no waiting in line
  • Email support

Max

The complete plan, with the most powerful model in the world.

US$41.10/mo

  • Everything in Pro, unrestricted
  • Claude Fable 5.2. the most advanced AI available

Enterprise

Authorized proxying and reselling. high-volume API use without ban risk.

US$130.00/mo

  • Everything in Max, unlimited
  • GPT-6 Astra. Enterprise exclusive
  • 1M context window on all models
  • Maximum response speed
  • Priority access during peak hours
  • Use on multiple devices at once
  • VIP support
  • Maximum queue priority and maximum speed
  • 1,000 RPM. built for heavy demand
  • Proxying and reselling authorized on top of the service. exclusive to this plan
  • Share with your company or clients. no high-volume ban
  • Up to 50 concurrent IPs per key
  • Instant priority support, 24/7

LumiaRouter For Teams

For companies with massive demand: 15,000 RPM, top of the queue and the developers one ticket away.

US$295.00/mo

  • Everything in Enterprise, unlimited
  • 15,000 RPM. massive demand
  • Top queue priority on every model
  • Full priority support: instant response by ticket
  • Direct contact with the LumiaRouter developers
  • Up to 200 concurrent IPs per key
  • Upgrade any time, paying only the difference

Real reviews, real customersReal reviews, real customers

Customer reviews from around the world — LumiaRouter started on Gamsgo, and our community's feedback speaks for itself, with over 400 positive reviews from real customers.

Discord #customer-chat: bubbadk thanks LumiaRouter for the speed being back while building a farm game
“Just to give some good vibe. I’m impressed and satisfied. The speed is up again and I’m building a farm game. So thank you lumiarouter ❤️”
bubbadkDiscord · #customer-chat
WhatsApp message about savings with Claude Code
“Seriously, the amount of money im saving and making with that Claude Code from Ghostcli subscription is unreal. The support team is super fast.”
EthanWhatsApp · lumiarouter.com Network
WhatsApp customer group: a new member asks if the service is legit
“Is this service legit? — Testing whole day and it’s a real deal.”
Souhail and KonradWhatsApp · lumiarouter.com customers
WhatsApp customer group: a sceptical member is answered by a paying customer
“Idk man, seems scammy. — I had the same thoughts at first, but I gave it a shot — trust me, it’s banging.”
Souhail and KonradWhatsApp · lumiarouter.com customers
Grupo de WhatsApp lumiarouter.com Network: Miguel M. diz que o projeto está rodando e Pedro diz que o Claude já imprimiu $100.00 USD (about $100.00 USD) de um cliente.
“O Claude já imprimiu $100.00 USD de um cliente.”

Claude already made $100.00 USD (about US$100) for one client.

PedroWhatsApp · lumiarouter.com Network
Victor Petry manda "tá a mil aqui" e a foto do próprio monitor com o chat da LumiaRouter aberto no Claude Fable 5.1.
“Tá a mil aqui.”

Going full speed here.

Victor PetryWhatsApp · lumiarouter.com customers
Souza responde a quem perguntou se alguém já usou por um mês, mostrando 9954 mensagens no período e 1426 num único dia.
“9,954 messages sent in one billing period, 1,426 of them in a single day — screenshot from the customer group.”
SouzaWhatsApp · lumiarouter.com customers
QUESTIONS

Frequently asked

Because these ten are the ones that hold up on real code. Claude Opus 5.5, Sonnet 5, Sonnet 5.5 and Fable 5.2 lead on agentic work; GPT-6 Astra, GLM 5.3, GLM 5.3 Flash, Hunyuan Hy4, DeepSeek V4.1 and MiMo V2.6 cover the rest, from long context to the lowest cost per token. All on the same key, and you pick per task.

Start now, no card needed

Create the account in under a minute, and subscribe only when it makes sense.

Discord

Still have questions?

Telegram Support
Discord