Disclosure: Sandy Music Lab is reader-supported. Some links on this page are affiliate links, and as an Amazon Associate we earn from qualifying purchases at no extra cost to you.
Neural Amp Modeler (NAM) is a free, open-source system that uses neural networks to capture real amps and pedals. To use it, install a free NAM-compatible plugin, download a capture (a .nam file) from TONE3000, add a cabinet IR if the capture is amp-only, and play through an audio interface.
NAM has gone from a hobbyist project to the most talked-about tone technology in guitar. It’s the open-source corner of the capture world we cover in our guitar amp modeling guide, and since June 2026 its new A2 architecture has started showing up in hardware too.
This guide shows you how to use neural amp modeler tones for free on your computer, what A2 changes, where to find the best profiles, and how to capture your own amp. If you’re weighing NAM against commercial options, read our NAM vs ToneX vs Quad Cortex comparison next.
🎸 Free download
Not sure which guitar is right for you?
Get the free Guitar Gear Cheat Sheet: match your style, budget and hands to the right guitar in 2 minutes. You’ll also get my 30-Day Practice Plan as a bonus.
Send My Cheat Sheet →Table of Contents
Quick Picks at a Glance
| Pick | Who it’s for | Skip it if | |
|---|---|---|---|
| Focusrite Scarlett Solo 4th Gen | The simplest interface for NAM at home | You need two mic inputs | Check price on Amazon |
| Universal Audio Volt 1 | Players who want a vintage-style preamp option | You want the smallest possible budget | Check price on Amazon |
| Valeton GP-5 | A cheap pedal that loads NAM-based tones | You want native A2 playback | Check price on Amazon |
| HeadRush Prime | Gigging players who want native NAM A2 on the floor | You only play at home | Check price on Amazon |
What Is Neural Amp Modeler?
Neural Amp Modeler was created by Steve Atkinson, PhD, and released as open-source software. Instead of simulating an amp’s circuit, NAM trains a neural network on recordings of a test signal played through the real gear. The result is a small .nam file that reproduces that amp, pedal or full rig at the settings used during the capture.
Because it’s open source, anyone can make captures, and any plugin or hardware company can support the format. That openness is why TONE3000 (the community library formerly known as ToneHunt) now hosts a huge, free library of captures. You can read the project’s own documentation at neuralampmodeler.com.
How to Use Neural Amp Modeler Free: Step by Step
- Connect an audio interface. Plug your guitar into the interface’s Hi-Z or instrument input. Laptop headphone jacks and USB guitar cables add too much latency.
- Install a NAM plugin. Download the free official NAM plugin from neuralampmodeler.com, or the free TONE3000 plugin, which runs A2 natively and browses the library from inside the plugin. Both run as VST3 or AU inside a DAW, and the official plugin also runs as a standalone app.
- Set your buffer. In your DAW or the standalone app, choose 48 kHz and a buffer of 64–128 samples for low latency. Our recording latency guide explains the trade-offs.
- Download a capture. Find a tone on TONE3000 and download the .nam file, choosing the A2 version where available.
- Load the capture and an IR. Load the .nam file in the plugin’s model slot. If it’s an amp-only capture, load a cabinet impulse response in the IR slot. Our guide to guitar IRs lists free options.
- Gain-stage it. Set the input so your hardest strums peak around −12 to −6 dBFS, and use the plugin’s input knob to match the level the capture expects.
- Add a gate and EQ. Use the built-in noise gate for high-gain tones and the EQ to fit the mix. Then add delay and reverb plugins after NAM.
Neural Amp Modeler A2: What Changed in 2026
Neural amp modeler A2 (Architecture 2) launched on June 2, 2026. It was built by TONE3000 in partnership with Steve Atkinson and is now the default for new captures on TONE3000. The key changes:
- One model, two sizes. A2 is “slimmable”: a single file runs as A2-Full for computers and A2-Lite for low-power pedals and amps.
- Better efficiency. TONE3000 reports that A2-Full uses about 30–40% less CPU than A1-Standard, and that A2-Lite runs at about 50% CPU on a 600 MHz ARM Cortex-M7, a chip that costs around $3 in bulk.
- Higher accuracy. When TONE3000 retrained its library, the median error-to-signal ratio (ESR) roughly halved, from 0.0062 with A1-Standard to 0.0033 with A2-Full.
- Longer memory. A2’s receptive field is about 132 ms at 48 kHz, up from about 85 ms, so it captures sag and compression better.
- Not backward-compatible. Plugins and hardware need an update (NeuralAmpModelerCore v0.5.2 or later) to play A2 files. A1 files still work where A1 is supported.
| Feature | NAM A1 (Standard) | NAM A2-Full | NAM A2-Lite |
|---|---|---|---|
| Released | Before 2026 | June 2026 | June 2026 (same file as A2-Full) |
| Typical use | Plugins | Plugins, powerful hardware | Budget pedals and amps |
| CPU vs A1-Standard | Baseline | About 30–40% less | Far less |
| Median ESR (TONE3000 library) | 0.0062 | 0.0033 | Close to A1-Standard |
| Receptive field | ~85 ms | ~132 ms | ~132 ms |
Full technical details are in TONE3000’s A2 guide. Keep in mind that TONE3000 ran the accuracy and listening tests itself.
Neural Amp Modeler Profiles: Where to Find the Best Tones
“Profiles” or “captures” are the .nam files you load. Most neural amp modeler profiles come from TONE3000, where you can filter by gear type (amp, amp and cab, pedal, full rig), by architecture (A2 or A1 legacy) and by popularity.
Tips for Choosing Profiles
- Check the capture type. “Amp + Cab” captures are ready to play. “Amp” or “DI” captures need an IR.
- Grab a pack, not a single file. Each capture is one amp at one setting. Packs give you clean, crunch and lead versions of the same amp.
- Look for A2 badges. TONE3000 has retrained most of its popular A1 library to A2.
- Read the notes. Good creators list the amp settings, the boost pedal used and the recommended input level.
- Try pedal capt