DeepSeek-V4.1-Flash · OpenAI-compatible API

DeepSeek models,
paid in Rupiah.

One endpoint that works with the OpenAI SDKs you already use. Top up a prepaid balance with QRIS from any Indonesian banking or e-wallet app — no foreign credit card, no subscription.

Get started Read the docs
Input · cache hit
$0.010
Input · cache miss
$0.400
Output
$1.600

USD per 1M tokens on deepseek-flash · same price every hour · Pricing

from openai import OpenAI client = OpenAI(    base_url="https://api.larasrouter.com/v1",    api_key="lr_live_••••",) response = client.chat.completions.create(    model="deepseek-flash",    messages=[{"role": "user", "content": "Halo, Laras Router!"}],) print(response.choices[0].message.content)
Two highlighted lines are the whole migration: the base URL and your key.

How it works

From QRIS to first completion in three steps.

  1. 1

    Top up with QRIS

    Pick any amount from $1 to $100. The invoice shows the exchange rate and the IDR total before you scan it with your banking or e-wallet app.

    $1 – $100 · paid in IDR
  2. 2

    Create an API key

    Name a key in the console. It is shown once, so store it like a password. Each key allows 300 requests per minute and 16 at a time.

    lr_live_••••••••
  3. 3

    Call the API

    Point any OpenAI SDK at our base URL. Streaming, tool calls, and JSON output work the way your code already expects.

    POST /v1/chat/completions

The console

Top up in Rupiah. Track every token in dollars.

This is the real Laras Router console, shown with example data. Your balance, keys, and every request live in one place, and the QRIS invoice tells you the exact Rupiah total before you scan.

console.larasrouter.comExample data
Laras Router console overview for an example workspace: available balance, spend and requests for the last 7 days, active API keys, recent requests with their cost, and a quickstart showing https://api.larasrouter.com/v1.Laras Router console overview for an example workspace: available balance, spend and requests for the last 7 days, active API keys, recent requests with their cost, and a quickstart showing https://api.larasrouter.com/v1.

Screens from the console with a fictional workspace. Amounts follow the published prices.

Integrations

Works with the tools you already use

Any agent, editor, or SDK that lets you set an OpenAI-compatible or Anthropic base URL can use Laras Router. Each guide is checked against the tool’s own documentation.

See all integrations

Claude Code is supported. Codex support is in development.

Agents

Editors & extensions

SDKs

Pricing

Two models. Published per-token prices.

You are charged for the tokens a request actually used, at the rates on this page. No tiers, no time-of-day pricing, no monthly fee.

  • Prepaid USD balance, topped up in IDR via QRIS
  • Credits never expire and are used only for API calls
  • Charges shown to six decimal places in your console
Pricing and cost calculator

DeepSeek-V4.1-Flash

1M context · 384K max output

deepseek-flash
Input · cache hit
$0.010/ 1M tokens
Input · cache miss
$0.400/ 1M tokens
Output
$1.600/ 1M tokens

DeepSeek-V4-Pro

1M context · 384K max output

deepseek-v4-pro
Input · cache hit
$0.060/ 1M tokens
Input · cache miss
$1.760/ 1M tokens
Output
$5.280/ 1M tokens

Worked example: one request

Cost of one example request
TokensCountDeepSeek-V4.1-FlashDeepSeek-V4-Pro
Cache-hit input8,000$0.000080$0.000480
Cache-miss input4,400$0.001760$0.007744
Output1,800$0.002880$0.009504
Charged$0.004720$0.017728

Why Laras Router

Built around how Indonesian developers pay and ship.

  • Pay in Rupiah with QRIS

    Any QRIS banking or e-wallet app works. No international credit card needed.

  • Prepaid, no subscription

    No card on file and no monthly minimum. Add balance when you need it and spend it per token.

  • Credits never expire

    Your balance is held in USD, so later exchange-rate moves never change it. Idle months cost nothing.

  • A drop-in for the OpenAI SDK

    Change the base URL and the key. The official Python and Node SDKs, and any client that accepts a base URL, keep working.

  • Every token accounted for

    Each request is metered as cache-hit input, cache-miss input, and output, and its cost is shown to $0.000001.

  • Prompts are not stored by us

    Prompt and completion text passes through our gateway in memory. It is sent to DeepSeek to generate the response.

Good to know before you start

Input
Text, plus images on models with vision
Per-key limits
300 requests / min · 16 concurrent
First request
Needs a top-up; there is no free credit
Spend control
Optional monthly budget per key; revoke a leaked key at once

Send your first request today.

Create an account, top up from $1 with QRIS, and call deepseek-flash with the SDK you already use.

Get started Read the docs

Questions? Our Indonesian support team is at [email protected].