Private beta, by invitation

Change one line.
Cut the bill by up to 92%.

One endpoint that speaks both the Anthropic and OpenAI APIs. Claude Code, Cline, the official SDKs: they all run without touching your code. Prepaid balance in USD, drawn per token, no card on file.

Invite-only until automated top-ups are finished.

request
POST /v1/messages
model
claude-opus-5
input
12 400 tokens
output
1 850 tokens
LajuAPI$0.0119
List price$0.1082
89%cheaper for exactly the same request

Worked out from the rates in the table below. An example, not a real invoice.

... Uptime 24h... Latency... Checked... Full status →
Scroll ↓
Pricing

Pay per token, in USD.

Prepaid balance, drawn down on every request. Every price below is per 1 million tokens.

ModelInputCached inputOutputDelta
Anthropic · Claude
claude-opus-5$5.00$0.55$25.00$2.75−89%
claude-opus-4-8$5.00$0.55cached $0.055$0.055$25.00$2.75−89%
claude-sonnet-5$3.00$0.25$15.00$1.25−92%
claude-haiku-4-5$1.00$0.12$5.00$0.60−88%
claude-fable-5-1$10.00$5.00$50.00$25.00−50%
OpenAI · GPT
gpt-5.6-sol$5.00$0.40$30.00$2.50−92%
gpt-5.6-terra$2.00$0.20$12.00$1.25−90%
gpt-5.6-luna$0.20$0.04cached $0.004$0.004$1.20$0.25−80%
gpt-6-astra$10.00$2.00$50.00$10.00−80%
Moonshot · Kimi
kimi-k3$3.00$2.00cached $0.20$0.20$15.00$10.00−33%
xAI · Grok
grok-4.6$2.00$0.30cached $0.03$0.03$6.00$0.90−85%
grok-4.5$2.00$0.30$6.00$0.90−85%
Z.ai · GLM
glm-5.3$1.40$1.00$4.40$3.20−29%
glm-5.3-flash$0.15$0.12$0.50$0.40−20%
Google · Gemini
gemini-3.1-pro-preview$2.00$0.50$12.00$3.00−75%
gemini-3.7-flash$0.75$0.45$3.75$2.70−40%
gemini-3.8-flash$0.75$0.45$3.75$2.70−40%
gemini-3.1-flash-lite$0.25$0.125$1.50$0.75−50%
DeepSeek
deepseek-v4-pro$0.66$0.25cached $0.025$0.025$1.98$0.50−62%
deepseek-v4-flash-0731$0.22$0.10$0.66$0.20−55%
Alibaba · Qwen
qwen3.8-max$2.00$1.00$6.00$3.00−50%

$0.00List price

Only models with a price in the Cached input column support prompt caching. Models showing an em dash bill every input token at the standard rate. Caching needs a stable prefix of at least 1,024 tokens; the cached rate applies from the second request that reuses it.

Work it out yourself

Drag to your own monthly volume. The comparison assumes 4 input tokens for every 1 output token, which is roughly what coding traffic looks like.

...
LajuAPI$0.00
List price$0.00
0% ...

Images, billed per image

ModelOutputPer image
Images
gpt-image-2-1k≈1.6 MP$0.12
gpt-image-2-2k≈4.2 MP · 2048²$0.20
gpt-image-2-4k≈14.7 MP · 3840²$0.30
gemini-3-pro-image1024² – 2048², your choice$0.08

Billed per image rather than per token. Each tier fixes the output pixel budget (aspect ratio can still follow the prompt), and the size and quality parameters are ignored: the tier is the thing you buy. No provider list price is shown, because providers publish image rates per quality and size rather than per these tiers.

See full pricing

Every model is served under its own name and nothing is relabelled in between. Your balance never expires. Top up with USDT, or in Indonesia with QRIS, e-wallet, or bank transfer; during the beta we issue top-ups by hand as redemption codes. Beta rates track the aggregator market and can move, but balance you already bought is honoured at the rate you paid. “List price” is each provider's own published rate as of July 2026, shown for comparison only.

Models

Every model runs under its own name.

Nothing is quietly swapped for something cheaper halfway through. What you call is what runs, and we test our own suppliers on a schedule to keep it that way.

Quickstart

Two lines of config.

Change the base URL and the key. Streaming and tool use pass through untouched, which is what makes agentic tools usable here.

# point Claude Code at LajuAPI
export ANTHROPIC_BASE_URL=https://api.lajuapi.com
export ANTHROPIC_AUTH_TOKEN=sk-laju-••••••••

claude
Claude CodeopencodeCline Roo CodeContinueKilo Code Copilot BYOKAnthropic SDKOpenAI SDK LangChaincurl

Any client that lets you set a custom base URL works.

What we promise

The honest version.

Cheap access is easy to promise and easy to get wrong. Here is what we commit to, and what we do not.

Availability

Best-effort, no SLA

Several suppliers sit behind each model with automatic failover. The status figures above come from a real monitor, not from decoration. We do not promise uptime we cannot back.

Integrity

No silent model substitution

You get the model you asked for. We run scheduled checks against our own suppliers for exactly this, and a lane caught pretending is dropped from rotation.

Billing

Prepaid, full stop

No stored card, no surprise invoice at the end of the month. When the balance runs out, requests stop.

Failure modes

Errors you can predict

Every failure comes back as a documented error with a request ID. Raw upstream messages are never handed to you.

Data

Honest about logging

Request metadata is kept for billing. Prompt bodies stay only for a short debugging window. The infrastructure behind us logs on its own terms, so we will not claim zero logging.

Standing

Independent service

We are nobody's official reseller and are not affiliated with any model provider. What we offer is compatibility with their API formats.

FAQ

Questions worth asking.

Is this an official Anthropic or OpenAI service?

No. LajuAPI is an independent gateway that speaks the same API formats, so the SDKs and coding tools you already use work unchanged. No affiliation, no endorsement.

How can it be this much cheaper?

We buy capacity in bulk through the aggregator market and resell it at a margin, on our own infrastructure and our own keys. That is the whole trick: no special deal implied, no partnership claimed. Our pricing tracks that market, which is why rates here are indicative and why the rate you already paid is honoured.

Do streaming and tool use work?

Yes. Both pass through unchanged, which is what makes agentic tools like Claude Code usable here. Whatever your client sends in the Anthropic or OpenAI request format is forwarded as it is; the only things you change are the base URL and the key.

How do I pay?

Prepaid balance in USD. Internationally that means USDT; in Indonesia, QRIS, e-wallet, or bank transfer also work. During the beta you pay first and we issue a redemption code you apply to your balance. Automated top-ups land with general availability.

Are my prompts stored?

We keep request metadata: timestamp, model, token counts, latency, request ID, for billing and abuse handling. Prompt and completion bodies are kept only for a short debugging window. Your requests are processed by third-party model infrastructure whose retention policies we do not control.

Can I get an account now?

Invite-only while the billing flow is finished. The API, dashboard, and status page are already live, and public signup opens together with automated top-ups.

Point your tools at one endpoint.

The API, dashboard, and status page are live. Signup opens with automated billing.