Can I keep the OpenAI SDK?

Yes. Set base_url to https://api.scalattice.cloud/v1 and use a Scalattice API key. chat.completions and models work today. Embeddings are not enabled yet.

How do I pin a region or vet outputs?

Send policy with the request. X-Scalattice-Region pins EU, US, or APAC. Output vetting runs independent checks before a result returns. tier2.5 splits work across operators.

How is inference priced?

Published per-token catalog rates with prepaid credits. Compare your current host with the savings calculator on this page. Model-specific promos run on Scalattice Cloud.

Developers

What are you trying to ship?

Then we will not ask you to install a new client. The next question is whether that SDK can stay untouched.

Then we start from what has to be true in production, not from a feature list.

Then tell us what usually ends the conversation. We will answer that first.

What does production inference have to include?

Swap the base URL and the key. The client you already run in production stays.

Those ride on the request. Scalattice assigns operator capacity against that policy.

Catalog rates, prepaid credits. No contract negotiation to make a first call. The savings calculator turns that into a monthly number.

Should the client stay as it is?

Point the OpenAI client you already have at Scalattice. Same chat.completions calls. Embeddings are not enabled yet.

from openai import OpenAI

client = OpenAI(
    base_url="https://api.scalattice.cloud/v1",
    api_key="slt_…",
)

client.chat.completions.create(
    model="qwen-3-8b",
    messages=[{"role": "user", "content": "Ship it."}],
)

Two environment variables. Any OpenAI-compatible library picks them up.

OPENAI_API_KEY=slt_…
OPENAI_BASE_URL=https://api.scalattice.cloud/v1

What has to be true first?

Per-token catalog pricing with prepaid credits. Model-specific promos live on Scalattice Cloud. Use the savings calculator if you want a monthly number. Nothing is quoted behind a sales call.

Output vetting runs independent checks and compares them before anything returns to your app. It is a request policy, not a separate product.

Pin EU, US, or APAC with a request header. Scalattice routes operator capacity inside that boundary.

What usually kills a vendor for you?

Rates are published. You prepay credits. See pricing or the savings calculator before you create a key.

There is not one. Python, Node, curl, or any OpenAI-compatible client. You change the base URL and the key.

Region, vetting, and security tiers travel with the request. The next question is which of those you actually need.

This is the only switch.

Create a token in Scalattice Cloud. Use it as the OpenAI API key. Point the client at https://api.scalattice.cloud/v1.

from openai import OpenAI

client = OpenAI(
    base_url="https://api.scalattice.cloud/v1",
    api_key="slt_…",
)

The last card is the door. Keys, pricing, and the CLI sit there.

Then we ask what has to ride along: region, vetting, or a security split.

What has to ride with every request?

Send X-Scalattice-Region: eu, us, ap, or auto. Sensitive work stays inside that pool. How the lattice routes.

Independent checks run on the result. You compare them before your app sees a completion.

tier2.5 splits the job across operators. Stronger security policy, same SDK.

Then the request is model plus messages. Routing still happens. You just do not set a policy header.

Your next step is a key, not a meeting.

Scalattice is a managed inference API. You set model, region, vetting, and security policy. Scalattice routes operator capacity, verifies outputs, and bills per token.

Get API keys Estimate savings

Client: OpenAI SDK drop-in. Python, Node, curl.
Calls: chat.completions and models. No embeddings yet.
Policy: Region, output vetting, tier2.5 on the request.
Price: Published per-token rates, prepaid credits.

Models on the network

Qwen

Llama

DeepSeek

Gemma

Mistral

gpt-oss

Bill

What would inference cost you?

Enter your current host if you have one.
We pull live list prices for the same open models.

Your usage Pick models, optional current provider, volume, and mix.

Current provider

Optional. Leave blank to see what we would charge.

Catalog models

Full catalog. Toggle any mix.

Monthly tokens 20M tok Prompt plus completion tokens across the models you selected.

Output share 40% out Chat leans output. RAG and long context lean input.

Typical workloads Jump to a common inference shape.

Scalattice rates are live from the catalog. Rival rates are the latest public list prices we snapshot for the same open models. Monthly tokens are split evenly across selected models. Actual bills vary with routing, promos, retries, and token mix, and are not guaranteed. Figures are illustrative only.

$0 Estimated monthly cost

Estimating…

Get API keys

Developers

Production inference you control.

Same OpenAI SDK. Built-in output vetting, regional policy, and security tiers on every request.

Get API keys Estimate savings

Output vetting
Regional routing
Published pricing

Quick start

Point your client here.

from openai import OpenAI

client = OpenAI(
    base_url="https://api.scalattice.cloud/v1",
    api_key="slt_…",
)

client.chat.completions.create(
    model="qwen-3-8b",
    messages=[{"role": "user", "content": "Ship it."}],
)

Three lines to your first request

Platform

From API key to production.

Docs, dashboards, and predictable pricing so your team ships instead of negotiating contracts.

Cost

Published rates

Per-token catalog pricing with prepaid credits. Model-specific promos run on Scalattice Cloud.

Quality

Built-in vetting

Run multiple independent checks and compare outputs before anything returns to your app.

Regions

Data control

Limit requests to the geography your compliance team needs.

# .env: point any OpenAI client here
OPENAI_API_KEY=slt_…
OPENAI_BASE_URL=https://api.scalattice.cloud/v1
# Optional: X-Scalattice-Region: eu | us | ap | auto

# Same client, lower invoice

Scalattice Cloud: keys, usage, billing

Integration

Plug and play with OpenAI.

No new library to install. Set your API token and base URL, then use the OpenAI client you already run in production.

Token

API key

Create a Scalattice token in Scalattice Cloud and use it as your OpenAI API key.

Base URL

One endpoint

Point OPENAI_BASE_URL or base_url at https://api.scalattice.cloud/v1.

Client

Keep your stack

Python, Node, curl, or any OpenAI-compatible library works unchanged.

Models

Same calls

chat.completions and models today. Embeddings are not enabled yet.

Works with models you already know

Qwen

Llama

DeepSeek

Gemma

Mistral

gpt-oss

Routing

Set routing policy per request.

Region, vetting, and security policies travel with each call. Scalattice assigns capacity automatically. Learn about the lattice.

Geo routing

Pin sensitive workloads to EU, US, or APAC pools with a request header.

Vetting

Run multiple checks and compare outputs before returning a result.

Capacity fallback

Scalattice can fill shortfalls when provider GPUs are unavailable.

Distributed infrastructure for per-request routing — Per-request routing policies

Platform

Built for production inference.

Scalattice is a managed inference API. Set model, region, vetting, and security policy on each request. Scalattice routes operator capacity, verifies outputs, and bills per token.

OpenAI SDK drop-in

Swap base URL and API key. Use the client you already run in production.

Output vetting

Run multiple independent checks and compare outputs before anything returns to your app.

Regional policy

Pin workloads to EU, US, or APAC pools with X-Scalattice-Region.

Published per-token pricing

Catalog rates plus prepaid credits. Active promos and model grants on Scalattice Cloud.

Managed routing

Scalattice assigns operator capacity for each request based on your policy.

Security tiers

tier2.5 splits work across operators for stronger security policy.