TokForge Guides

Everything you need to get the most out of TokForge. From your first offline AI conversation to creating AI personas, generating images, and building automated agents with 150+ API endpoints.

Run this on your own phone

TokForge is free. Android is live on Google Play; iPhone and iPad are in public beta on TestFlight. No account, and it works with the network off once a model is downloaded.

Get it on Google Play iPhone beta on TestFlight

Reading on a computer? Check which models fit your phone, or see real measured speeds on the phone-speed leaderboard.

Start here

Which AI models run on your phone: RAM tiers, GGUF vs MNN, and quantization on mobile
Can you use ChatGPT offline?: what an on-device model can and cannot do
Can my phone run it: the RAM calculator, with its error bars stated
MNN vs GGUF on Android: which format, and which models exist in each
Phone-speed leaderboard: measured results from real handsets

Getting Started

How to Run AI Offline on Your Phone

Install TokForge, download your first model, and start chatting, all without internet.

Best AI Models for Your Phone

Which model fits your RAM? A visual guide from 0.8B to 27B.

Is Your Phone Fast Enough?

Check if your Android device can run local AI, and what to expect.

Features & Setup

Setting Up Character Cards

Import TavernAI characters, customize personalities, and build persistent AI companions.

Chat With Your Documents

Attach PDFs, Word docs, and EPUBs. Ask questions, get answers grounded in your files.

Speculative Decoding

Draft-verified decoding that speeds up structured output like counting, code, and JSON, plus when it helps versus when it doesn't.

TurboQuant: Leaner Context Memory

TQ4 aggressive attention-cache quantization shrinks context memory on small models. Here's how.

Performance

AutoForge & ForgeLab

One-tap auto-tuning that profiles your hardware and picks the fastest config.

Vulkan vs OpenCL vs CPU

Five GPU acceleration paths explained. Which one your chipset uses and why.

Developer & API

API Walkthrough

Enable the API, connect from your computer, and control TokForge with curl.

Building an AI Agent

Python scripts that orchestrate TokForge remotely: models, conversations, memory, and telemetry.

Comparisons

TokForge vs AnythingLLM vs Layla

An honest feature comparison of the top offline AI apps for Android.

9 Offline AI Apps, Compared on the Record

Engine, price, licence, install base and last update for nine apps, read off their own store listings.

MNN vs GGUF on Android

What each format is, which models exist in which, and why this page carries no speed multiplier.