NEWOrchestrate Claude Code, Codex and Warp Agent in Studio

Your own ChatGPT.Private, offline, yours.

Langtrain turns open-source models like Llama, Qwen and Mistral into an AI assistant trained on your company's data. It runs on your own hardware, even with no internet, and the weights are yours to keep.

Start training free See how it works

Runs fully offline
You own the model weights
Works with the OpenAI SDK
No ML team needed

trainingllama-3.1-8bLoRA · r=16

EXAMPLE RUN

TRAIN LOSS

0.590

STEP: 0520 / 1200
EPOCH: 1.30 / 3

step 0050loss 1.648 lr 2.0e-4
step 0100loss 1.403 lr 2.0e-4
step 0150loss 1.207 lr 1.9e-4
step 0200loss 1.083 lr 1.9e-4
step 0250loss 1.007 lr 1.8e-4
step 0300loss 0.920 lr 1.7e-4
step 0350loss 0.802 lr 1.6e-4
step 0400loss 0.693 lr 1.5e-4
step 0450loss 0.631 lr 1.4e-4
step 0500loss 0.599 lr 1.3e-4

NEWOrchestrate Claude Code, Codex and Warp Agent in Studio

Your own ChatGPT.Private, offline, yours.

Start training free See how it works

Runs fully offline
You own the model weights
Works with the OpenAI SDK
No ML team needed

trainingllama-3.1-8bLoRA · r=16

EXAMPLE RUN

TRAIN LOSS

0.590

STEP: 0520 / 1200
EPOCH: 1.30 / 3

step 0050loss 1.648 lr 2.0e-4
step 0100loss 1.403 lr 2.0e-4
step 0150loss 1.207 lr 1.9e-4
step 0200loss 1.083 lr 1.9e-4
step 0250loss 1.007 lr 1.8e-4
step 0300loss 0.920 lr 1.7e-4
step 0350loss 0.802 lr 1.6e-4
step 0400loss 0.693 lr 1.5e-4
step 0450loss 0.631 lr 1.4e-4
step 0500loss 0.599 lr 1.3e-4

Why Langtrain

Why not just use ChatGPT or Claude? Because it isn't yours.

Hosted assistants are a good start. They stop working for you the moment your data is sensitive, your domain is specialised, or your costs have to be predictable.

Your data leaves the building
Every prompt, document and customer record you paste in is sent to someone else's servers.
LANGTRAINA Langtrain model runs where your data already is. Nothing is sent out.
It doesn't know your business
General models learned the internet, not your products, policies, tickets or tone of voice.
LANGTRAINFine-tuned on your own documents and conversations, it answers the way your team would.
You rent it forever
You pay per token for as long as you use it, and the model can change or be retired without asking you.
LANGTRAINYou keep the weights. Self-hosted, there are no per-token fees and the model only changes when you retrain it.
It needs the internet
No connection, no assistant: not on a plane, a factory floor or a locked-down network.
LANGTRAINExport to GGUF and run it with llama.cpp on a laptop or server that never goes online.

How it works

Five steps to your own AI. No ML team required.

Upload what your team already has: help articles, tickets, chats, documents. Langtrain turns it into training data, trains an open model on it, checks it, and hands it back to run wherever you want.

Read the quick start

01DataUpload chat-format JSONL. One conversation per line.
02TrainLoRA, QLoRA, SFT, full fine-tuning or RLHF.
03EvaluateScore the fine-tune on held-out data before it ships.
04DeployLangtrain Cloud, Docker, or Kubernetes with the Helm chart.
05APICall it like any OpenAI-compatible model.

Open models, trained with our recipe. The hard parts are handled.

A base model like Llama or Qwen is a generalist. Our training pipeline makes it a specialist in your business: memory-efficient fine-tuning so big models fit small GPUs, preference tuning so it sounds like you, and automatic search so the settings are right the first time.

Runs offline

Your data never leaves. Neither does your model.

Training runs in an isolated environment with no inbound traffic. The result is a file you own: run it offline on a laptop, on your own servers, in your cluster, or on our cloud.

Training runs on dedicated, ephemeral instances in an isolated VPC with no ingress routing.

Isolated compute

Langtrain Cloud

Push the trained model with one command. It becomes available through the OpenAI-compatible API.

$ trainer.push("my-assistant")

Interface: OpenAI-compatible API
Deploy: One command

Deployment docs

Use cases

Built for teams whose data can't leave. One platform, every department.

All use cases

Compare

The control of building it yourself. The ease of an API.

Langtrain compared with hosted AI APIs and building your own stack
Capability	Langtrain	ChatGPT / Claude APIs	Build it yourself
Runs on your own hardware	Yes	No	Yes
Works with no internet	Yes	No	Yes
You own the model weights	Yes	No	Yes
Trained on your company's data	Yes	Limited	Yes
No ML team or GPU cluster to run	Yes	Yes	No
OpenAI-compatible API	Yes	Their own	You build it
How you pay	Free to start, then a plan	Per token, for as long as you use it	GPUs and engineers

Hosted APIs change their offerings often; check their current terms.

Products

A desktop app, an SDK, a CLI and an API. Use whichever suits your team.

In code

Prefer code? The same five steps, in Python.

Every snippet is copied from the docs, so what you see is what you run.

training_data.jsonl

from /docs/quick-start

{"messages": [{"role": "user", "content": "What is Langtrain?"}, {"role": "assistant", "content": "It's a platform for fine-tuning LLMs."}]}
{"messages": [{"role": "user", "content": "How do I start?"}, {"role": "assistant", "content": "Use the LoRATrainer class!"}]}

Pricing

Start free. Pay as your model earns its keep.

Compare plans

Starter
Your first private model.
$0forever
- 5M training tokens
- Langtrain Studio (desktop)
- You own the weights
Start training free
Pro
For teams training and serving models.
$20per month · ₹1,499 in India
- 10 agents
- Langtrain Studio Pro
- 200M training tokens
- Browser agent sessions
See Pro
Enterprise
For data that must stay in your environment.
Customannual agreement
- VPC and on-premise deployment
- Dedicated GPU clusters
- SSO and audit logs
Talk to us

FAQ

Before you start. Straight answers.

Stop renting AI. Own it.

pip install langtrain-ai

Train your first private model free. Keep the weights, run it anywhere, and never send your data to someone else's AI again.

Start training free Read the quick start

Your own ChatGPT.Private, offline, yours.

Your own ChatGPT.Private, offline, yours.

Why not just use ChatGPT or Claude? Because it isn't yours.

Your data leaves the building

It doesn't know your business

You rent it forever

It needs the internet

Five steps to your own AI. No ML team required.

Open models, trained with our recipe. The hard parts are handled.

LoRA and QLoRA

Supervised fine-tuning

Full fine-tuning

Preference tuning

Automatic tuning

Your data never leaves. Neither does your model.

Langtrain Cloud

Start from open models. Leave with your own.

Built for teams whose data can't leave. One platform, every department.

The control of building it yourself. The ease of an API.

A desktop app, an SDK, a CLI and an API. Use whichever suits your team.

Prefer code? The same five steps, in Python.

Start free. Pay as your model earns its keep.

Starter

Pro

Enterprise

Before you start. Straight answers.

How is this different from ChatGPT or Claude?

Can it really run offline?

Which models can I start from?

Do I need ML expertise?

What data format should I use?

How much data do I need?

How long does training take?

Is my data private?

Can I export and self-host my model?

Stop renting AI. Own it.

Why not just use ChatGPT or Claude? Because it isn't yours.

Your data leaves the building

It doesn't know your business

You rent it forever

It needs the internet

Five steps to your own AI. No ML team required.

Open models, trained with our recipe. The hard parts are handled.

LoRA and QLoRA

Supervised fine-tuning

Full fine-tuning

Preference tuning

Automatic tuning

Your data never leaves. Neither does your model.

Langtrain Cloud

Start from open models. Leave with your own.

Built for teams whose data can't leave. One platform, every department.

The control of building it yourself. The ease of an API.

A desktop app, an SDK, a CLI and an API. Use whichever suits your team.

Prefer code? The same five steps, in Python.

Start free. Pay as your model earns its keep.

Starter

Pro

Enterprise

Before you start. Straight answers.

How is this different from ChatGPT or Claude?

Can it really run offline?

Which models can I start from?

Do I need ML expertise?

What data format should I use?

How much data do I need?

How long does training take?

Is my data private?

Can I export and self-host my model?

Stop renting AI. Own it.