GL.iNet OpenWrt Build Guide

This document explains how to build supported GL.iNet firmware images from the OpenWrt 25.12 source tree. Common source and feed preparation steps are shared by every model, while each model uses its own complete OpenWrt .config, DTS, and image profile.

Supported models

Model	Complete configuration	Target
GL-BE10000	gl_configs/gl-be10000.config	MediaTek Filogic
GL-BE14000	gl_configs/gl-be14000.config	MediaTek Filogic

Only models listed in this table have a complete configuration ready for a direct build. Additional models must provide a separate configuration named gl_configs/<model>.config before they are added here.

1. Clone the source code

Install the standard OpenWrt build dependencies for your Linux distribution, then clone the openwrt-25.12 branch:

git clone --branch openwrt-25.12 --single-branch \
  https://github.com/gl-inet/gl-openwrt.git
cd gl-openwrt

If the source tree has already been cloned, update it before building:

git checkout openwrt-25.12
git pull --ff-only

The repository only needs to be cloned once. The same source tree can build any model listed in the supported-model table.

2. Update and install the feeds

./scripts/feeds update -a
./scripts/feeds install -a

Run both commands again whenever feeds.conf.default or a feed revision is updated. Feed preparation is shared by all supported models.

3. Select a model and copy its complete configuration

Set MODEL to a configuration name from the supported-model table. For GL-BE14000:

MODEL=gl-be14000
test -f "gl_configs/${MODEL}.config"
cp "gl_configs/${MODEL}.config" .config
make defconfig

Each file under gl_configs/ is a full .config, not a configuration fragment. Do not enable multiple device profiles in one configuration.

The complete configurations select the following profiles:

Model	Target System	Subtarget	Target Profile
GL-BE10000	MediaTek Ralink ARM	Filogic 8x0	GL.iNet GL-BE10000
GL-BE14000	MediaTek Ralink ARM	Filogic 8x0	GL.iNet GL-BE14000

GL-BE14000 default network

The default LAN address is 192.168.1.1/24. Ports lan1 through lan8 belong to LAN, the dedicated eth2 port is the DHCP WAN, and sfp is the second DHCP WAN.

These defaults apply to a clean installation or after resetting the configuration. An upgrade that keeps settings preserves the existing network configuration.

4. Build the firmware

Use all available CPU cores for a normal build:

make -j"$(nproc)"

If the build fails, rebuild serially with verbose output to obtain the complete error message:

make -j1 V=s

The generated images are written under:

bin/targets/<target>/<subtarget>/

Both currently supported models use this output directory:

bin/targets/mediatek/filogic/

Switching models

To build another supported model in the same source tree, copy that model's complete configuration and normalize it before building:

# Replace this value with a model listed in the supported-model table.
MODEL=gl-be14000
test -f "gl_configs/${MODEL}.config"
cp "gl_configs/${MODEL}.config" .config
make defconfig
make -j"$(nproc)"

When switching to a different target or subtarget, clean target-specific build artifacts first:

make dirclean

Then copy the new model configuration and start the build. Models on the same target and subtarget normally do not require make dirclean.

Complete command sequence

For a clean build environment, set MODEL to either gl-be10000 or gl-be14000 and run:

git clone --branch openwrt-25.12 --single-branch \
  https://github.com/gl-inet/gl-openwrt.git
cd gl-openwrt
./scripts/feeds update -a
./scripts/feeds install -a
MODEL=gl-be14000
test -f "gl_configs/${MODEL}.config"
cp "gl_configs/${MODEL}.config" .config
make defconfig
make -j"$(nproc)"

Adding another model

Keep model support isolated and reproducible:

1. Add a separate DTS and image profile for the model.
2. Build and verify the model configuration.
3. Save the complete configuration as gl_configs/<model>.config.
4. Add the verified model to the supported-model table in this document.

Share a common .dtsi only when models genuinely use the same hardware nodes. Do not combine unrelated board definitions or multiple model profiles in one configuration.