Aleroy — Inference on your terms, in your boundaries

Dedicated inference for open and custom models.

A fixed serving profile: we do not change the model, revision, precision, quantisation, region, or data boundary without 180 days' notice. Zero retention by default. What you tested is what stays in production.

For production teams evaluating open and custom model inference.

The platform

One API. One contract. Every placement.

FAR Labs is a unified platform for high-performance inference. Ready-to-use and custom models reach you through a single API — running on our capacity, in your cloud, or on your own hardware. Each request is routed against your requirements for model, speed, cost, and processing location.

Choose per request

Choose catalogue or custom models per task through one API, without adding a second integration.

‍

Choose the placement

Run on FAR Labs capacity, customer cloud, or customer hardware, with routing limited to approved locations.

‍

Know where it ran

Review placement details during onboarding and operations so teams know how requests are handled.

‍

Also available

Sandboxes for agents

An isolated workspace an agent can create and control through an API: run code, work with files, install libraries, test applications, use a browser. We provide the execution environment — your application keeps the agent's logic.

Full environment

CPU and RAM by the hour, for agents that need a real machine to work in.

Code Interpreter

A simplified option billed per session, for straightforward execution tasks.

Persisted state

Keep the working state so an agent can resume later, charged on volume and duration.

Control

Fail closed, never substitute silently.

‍

Reserved throughput, dedicated deployments, or capacity in your own environment — priced against what you actually run. We confirm the numbers for your workload during onboarding.

Talk to us about your workload.

‍

Share your workload and constraints. We will confirm the supported model options, deployment path, pricing, and onboarding steps before production.

‍

From catalogue models to your own weights

Model APIs

Initial model families we support or can support on request. Final availability, pricing, and serving profiles are confirmed during onboarding.

‍

Custom model inference

Upload weights or a container; FAR Labs optimizes, deploys, and scales dedicated inference in the target environment.

‍

Continuously fine-tuned models

For models you continuously fine-tune, FAR Labs handles deployment and reliable inference operations from updated weights or containers.

‍

Sovereign

A national cluster, not a foreign dependency

For governments and operators of national AI infrastructure: existing data centres, your own GPUs, and partner clusters brought together into one managed inference federation.

One virtual cluster

Bring approved data centers, clouds, and GPU clusters together as one managed inference federation.

‍

Your boundaries hold

Placement, control, scaling, and failover are reviewed against the sovereignty boundaries you define.

‍

Accounted per organisation

Each organization receives separate access rights, limits, and usage accounting through one shared platform.

‍

Pricing

Priced against your workload, not a rate card.

Model

Provider

Status

Pricing

Serving profiles

Kimi K3

Moonshot

Supported

On request

At onboarding

DeepSeek V4.1 Flash

DeepSeek

Supported

On request

At onboarding

Qwen3.8 27B

Alibaba

Supported

On request

At onboarding

MiniMax M3

MiniMax

Supported

On request

At onboarding

Final availability, pricing, and serving profiles are confirmed during onboarding.

If you own GPUs

Turn idle capacity into an inference business

Launch under your own brand

Offer inference under your own brand. FAR Labs adapts the customer portal, deploys the platform components, and supports the operating model behind it.

Connect capacity, share revenue

Connect reserved GPU capacity to FAR Labs and share in the inference revenue generated on that capacity. Commercial terms are confirmed with each partner.

What we handle

FAR Labs supports platform operations, customer support, model updates, fine-tuning work, and joint sales where it helps the partnership.

Get started

Tell us where it has to run.

Whether you're building on models, sitting on idle GPUs, or standing up national infrastructure — the starting point is the same conversation about your workload and your boundaries.