Features Compare Releases Download Our Mission
Private Beta · Native macOS App

Orchestrate your entire LLM pipeline. Locally.

Stop stitching together CLI tools. LLMForge handles the full pipeline — from model download to on-device deployment in one native window. No terminal. No cloud. Just results.

Download Free See how it works
100% Local & private
MLX Apple Silicon native
GGUF App-ready export

The entire pipeline. One window.

Every step of the LLM workflow, from finding a model to shipping it on-device, lives inside LLMForge. No context switching. No config files.

Step 01

Browse & download from HuggingFace

Search the entire model hub. See architecture, size, and RAM requirement before you download. One click pulls it into your local workspace.

Phi-3 Mini Llama 3.2 Qwen 2.5 Gemma 3 Many more
Phi-3 Mini microsoft/phi-3-mini-4k · 3.8B params
2.2 GB
Llama 3.2 3B meta-llama/llama-3.2-3b-instruct
2.0 GB
Qwen 2.5 1.5B Qwen/qwen2.5-1.5b-instruct
1.0 GB
Gemma 3 1B google/gemma-3-1b-it
0.7 GB
Step 02

Curate training data without writing scripts

Import CSV/JSONL, label manually, or have a local model generate pairs you accept or reject. Always outputs clean Alpaca or ChatML, ready for MLX.

Import Manual Label AI-Assisted JSONL
// dataset.jsonl — 847 examples { "instruction": "Explain LoRA in one line", "input": "", "output": "LoRA adds small trainable matrices to frozen layers for efficient fine-tuning." } { "instruction": "What is quantization?", "input": "", "output": "Reducing model precision..." }
Step 03

Fine-tune on Apple Silicon

Runs natively on MLX — no CUDA, no cloud GPUs. Sessions persist across restarts, overfitting is caught live, and crash recovery picks up exactly where you left off.

MLX Native LoRA / QLoRA Live Loss Checkpoints Overfitting Detection Persisted Sessions
Epochs
3
Learning Rate
2e-4
LoRA Rank
16
RAM Status
✓ 16 GB
Epoch 2/3 · Step 142/20067%
Step 04

Quantize & export to GGUF or CoreML

Pick your quantization level. Balance file size against quality. One click converts and exports a ship-ready model you can drop into Xcode.

GGUF CoreML llama.cpp Xcode-ready
Q8_0 — Best quality7.2 GB
Q4_K_M — Recommended3.8 GB
Q2_K — Smallest2.1 GB
Step 05

Test side-by-side. Ship with confidence.

Same prompt, two models, simultaneous responses. Compare quality and speed. Save great outputs back to your dataset for the next training run.

A/B Compare tok/sec Feedback Loop
Base Model

LoRA is a method for adapting large language models using low-rank matrix decomposition techniques...

12.4 tok/sec
Fine-tuned ✓

LoRA injects small trainable rank-decomposition matrices alongside frozen weights, enabling efficient domain-specific adaptation.

11.8 tok/sec
Step 06

Serve locally. Test your apps instantly.

Spin up an OpenAI-compatible API from any fine-tuned model. SSE streaming with stop-generation support. A live metrics overlay tracks tokens/sec, latency, and token counts on every request.

Local API Server OpenAI Compatible SSE Streaming Response Metrics Stop Generation
API Live localhost:8080/v1/chat/completions 1 req
34.2tok/sec
47prompt
128completion
84mslatency
You
Explain what LoRA does in one sentence.
API · Fine-tuned Model
LoRA injects small trainable low-rank matrices into frozen layers, enabling efficient task-specific adaptation without full retraining.
curl http://localhost:8080/v1/chat/completions \ -H "Content-Type: application/json" \ -d '{"messages": [{"role": "user", "content": "Hello!"}], "max_tokens": 256, "temperature": 0.7}'
Step 07

Push adapters to HuggingFace. Share your work.

Upload fine-tuned adapters directly to a HuggingFace repository from inside the app — public or private. No CLI, no manual file wrangling. One click from your library.

HF Push Adapter Upload Public & Private Repos One-click
Repository
gokulnair2001/qwen2.5-lora-adapter
Visibility
Private
Public
qwen2.5-adapter.safetensors
128 MB · Ready to push
Push to HuggingFace

Why developers choose LLMForge.

Everything you need to go from idea to deployed model — without the overhead.

Built-in model library

Browse, search, and manage all your local models in one place. Import, tag, and track versions without leaving the app.

Local API server

Expose any model as an OpenAI-compatible endpoint. Test your apps against localhost — no cloud deployment or API keys.

Push to HuggingFace

Upload fine-tuned adapters directly to a HuggingFace repository from inside the app. Public or private — no CLI required.

Fully offline

Your data, models, and training never leave your machine. Private by default.

RAM-aware

Checks memory before every operation. Clear warnings, never silent OOM crashes.

Apple Silicon native

MLX for training. Metal for inference. Built specifically for M-series chips.

Your next model
starts here.

Download LLMForge and go from a HuggingFace model to a ship-ready GGUF in under an hour. Free.

Download for macOS
Free Under 200 MB No account required

Requires Apple Silicon (M1+) · macOS 26+ · 8 GB RAM