Skip to main content

Dev Station Technology

AI Development

Llama development services that start with the licence

We have no published Llama build. Our AI product calls a Claude-compatible API, and each Llama version carries its own licence and use policy that you should read before you build.

0Published builds on Llama
4Newest Llama family on Meta pages
3.3Newest Llama 3 release listed
Per versionEach Llama release has its own licence
What we build

What we would take on with Llama

A licence read first

The licence and use policy for the version you choose, checked before any build.

Models behind an API

A FastAPI service in front of a model you host.

Cost control

Token counts and cost reported per workspace, as in Sales OS.

Model swaps

A model name kept in settings so a change is configuration.

Architecture

Where Llama would sit

A hosted model in the slot our API call fills
Clients
React logo
React 19 appAccount briefs and grades
API
Python logo
FastAPIPython 3.13, Pydantic 2
Model
LLaMA logo
Llama modelWhere it would sit
Claude-compatible APISales OS, today
Delivery
Docker logo
Docker ComposeMulti-stage build

Illustration. A dashed outline marks where the technology would sit. We have no published build on it.

Families

The Llama families Meta lists

Llama on the Meta developer pages
Llama 4Natively multimodal models with a mixture-of-experts architecture. It has its own Llama 4 License.
Llama 3Includes 3.1, 3.2 and 3.3 across the instruction-tuned text models. Llama 3.3 has its own licence.
Llama 2The older family, listed after Llama 3.

Source: Meta developer pages, dev.meta.ai/llama, checked 5 November 2026. The site now leads with Meta Muse models. Read each licence and use policy yourself.

Builds

The closest builds we can show you

Sales OS account research product built by Dev Station
AI product, in-house

Sales OS

An engine that crawls a site and writes a graded brief through a Claude-compatible API, with cost and tokens reported per workspace.

40 accountsAbout 26 hours of reading a week replaced by 40 credits
Core development team built by Dev Station for a German cloud-native company
Dedicated team, Germany

Core development team

A chatbot project delivered at a fixed price, with GraphQL in the stack.

60+Developers the team grew to
Fit check

When Llama is the right call

Pick Llama when

  • Data must stay inside your own cloud
  • You can host and operate a model
  • Your lawyer has read the licence for the version you pick

Look elsewhere when

  • You need a published Llama build to inspect, which we do not have
  • A hosted model API already fits your cost and privacy needs
  • You cannot accept the licence terms
Stack

What Llama usually sits beside

Python logoPython
PyTorch logoPyTorch
Docker logoDocker
Kubernetes logoKubernetes
TensorFlow logoTensorFlow
AWS logoAWS
FAQs

What buyers ask first

Which Llama models exist?

Meta lists Llama 4, the Llama 3 family with 3.1, 3.2 and 3.3, and Llama 2. Its developer site now leads with the Muse family and keeps Llama in its own section.

Can we use Llama in a commercial product?

Each Llama version has its own licence and use policy, such as the Llama 4 License. We do not give legal advice. Read the terms for your version and ask your lawyer.

Do you have a published Llama build?

No. We say so on the first call and show the nearest work, which is Sales OS on a Claude-compatible API and a fixed-price chatbot project.

Who owns the code and the cloud accounts?

You do, from the first commit. Repositories and cloud accounts stay in your name.

Tell us what your model has to do

Thirty minutes on your data, your hosting and your licence needs. No sales pitch.

Let's Talk