Skip to content
Aevonix ResearchContact
Menu
Aevonix Research

Open source · AGPL-3.0-only

FlashAlignReAlign models instantly, on demand.

Apply alignment profiles per request. Use abliteration for approved users and channels, or keep a profile active across every routed request.

View on GitHub

01

Choose when to apply FlashAlign

Verified contacts is the default. It applies a profile to messages from an approved channel, account and contact.

Always On applies the configured profile to every request routed through FlashAlign, including tool and subagent requests. It overrides per-tool alignment settings and stays selected across restarts. The harness still decides who can send messages and which tools can run. Always On does not mark callers as verified. Requests fail if the configured profile cannot be applied.

Restricting abliteration to verified users and channels reduces the risk of an abliterated model being abused through malicious prompt injection. Tool outputs within an authorized request still rely on the harness's protections.

The starter profiles focus on refusal reduction. Future profiles could also increase refusals for public or higher-risk tasks.

02

Compatibility

FlashAlign works with supported base and quantized weights without permanently changing them. Run it directly, with Hermes or OpenClaw, or through your own harness. Managed inference runs on Linux with NVIDIA GPUs.

Five starter profiles are available.

Experimental test profiles based on Keys, OrcaRouter, Dealign, and MoonMath abliterated checkpoints. Fitting and runtime packaging by Aevonix. See the exact checkpoint versions.

ModelSupported engines
DeepSeek V4.1 Flash 1017vLLM, TensorFold
GLM 5.3 FullTensorFold
GLM 5.3 FlashTensorFold, SGLang
Qwen3.8-27BTensorFold, vLLM, SGLang, llama.cpp
Gemma 4 26B A4BTensorFold, vLLM, SGLang, llama.cpp
Harness and model compatibility

03

Get started

Download a supported model first.

git clone https://github.com/Aevonix/FlashAlign.gitcd FlashAlignbash ./install.sh --auto

The wizard detects local models and harnesses, selects a matching profile, prepares the engine and creates a deployment signing key.