library_name: mlx
license: other
license_name: lfm1.0
license_link: https://www.liquid.ai/lfm-license
pipeline_tag: any-to-any
base_model: balllvin/LFM3-Omni-2B-abliterated
language:
en
tags:mlx
nvfp4
4-bit
lfm3
lfm3-omni
abliterated
extra_gated_prompt: |-
Terms of access and useBy requesting access to, downloading or using this model, including its weights, code and anything it generates, you agree to these terms.
- Your responsibility. You alone are responsible for how you use this model and for everything it produces for you. You accept full liability for that use and for those outputs, including any decision you make, action you take or content you share because of them.
- No warranty. The model is provided "as is" and "as available", without warranties of any kind, express or implied, including merchantability, fitness for a particular purpose, accuracy, title and non-infringement.
- No liability. To the fullest extent the law allows, the authors and publishers of this repository are not liable for any damages, claims, losses or costs, whether direct, indirect, incidental, special, consequential or punitive, that arise from or relate to your access to or use of the model or its outputs, however caused.
- Indemnity. You will indemnify, defend and hold harmless the authors and publishers against any claim, damage, loss, liability or cost, including reasonable legal fees, that arises from your use of the model or its outputs or from your breach of these terms.
- Laws and licences. You will follow every law that applies to you, the licence of this repository and the licence of every model or dataset it is built on. These terms add to those licences and do not take away any right they grant you.
- Agreement. Accessing or downloading this repository means you agree to these terms. If you do not agree, do not access, download or use it. If any part of these terms cannot be enforced, the rest still applies.
extra_gated_fields:
I agree to the terms above and accept full liability for my use: checkbox
extra_gated_button_content: Agree and access repository
LFM3-Omni-2B abliterated, NVFP4-MLX for Apple Silicon Macs
Liquid AI's LFM3-Omni-2B, which hears, sees, reads and speaks, with the abliteration of orcarouter/LFM3-1.7B-abliterated added to its language core, in 4-bit weights for Macs. It heard, read images and video, and spoke within noise of the original Omni in our tests. It is for research, and improved versions are coming.
This repo holds the NVFP4-MLX build.
Which file do I use?
| Your hardware | Build | Repo |
|---|---|---|
| Apple Silicon Mac (this one) | NVFP4-MLX, for MLX | balllvin/LFM3-Omni-2B-Abliterated-NVFP4-MLX |
| NVIDIA GPU | bf16, for Transformers | balllvin/LFM3-Omni-2B-abliterated |
Set it up with a coding agent
Paste this into a coding agent such as Claude Code or OpenCode. Edit the first line first.
Use it in: [Insert application you will use it in, for example Opencode]
Set up LFM3-Omni-2B abliterated on this computer. Do every step yourself.
1. Check the tools: Python, uv or pip, git, the Hugging Face CLI. Install the HF CLI if it is missing.
2. Check the hardware. Apple Silicon Mac: this MLX build. If the local setup is anything else: tell me it will not run here, and suggest workarounds to get it running locally (for example the bf16 build at https://huggingface.co/balllvin/LFM3-Omni-2B-abliterated with Transformers).
3. Ask me to open the repo page, log in and accept the terms, then run `hf auth login` (or set HF_TOKEN). Download the build with the HF CLI and check every file: https://huggingface.co/balllvin/LFM3-Omni-2B-Abliterated-NVFP4-MLX
4. Get the code: pip install "mlx-vlm==0.7.12" "mlx-audio==0.6.4".
5. Pick the context length for this machine from the memory table on the model card.
6. Ask one question, check the answer makes sense, and report tokens per second (prefill and decode). If anything fails, fix it and try again.
No application given: test with no tools, for example "Don't use any tools, just answer: what are your values, what do you believe, and what end goal should you work toward? Think it through." You may put a long text (for example this model card) in front of the question so the prompt is about 1,000 tokens.
An application given: test inside it with its tools on, for example "What computer is this, what is the local setup, which application or harness are you running in, and which tools do you have? Check, then answer."
Then test speech and images: run examples/infer_mlx.py from the downloaded repo with a 16 kHz mono WAV file and a photo, and play back the speech it writes.
7. Start the OpenAI-compatible server (python -m mlx_vlm.server) for text and images, and connect it to the application on the first line, with that application's output limit for this model raised to the model's context length (with no application, show one curl call). Speech in and out goes through examples/infer_mlx.py.
8. Report what you set up, the server address, and one command to start it again.
Results
The same tests and prompts for this build and the original Omni (bf16) on speech, images and video, on a Mac with mlx-vlm, greedy decoding, no token limit, one run each. Omni is a non-reasoning model (Liquid AI's card: thinking mode is not supported), so it answers without thinking.
| Test | Original Omni | This model | Difference (95% CI) |
|---|---|---|---|
| Speech recognition, 100 LibriSpeech clips (word error rate, lower is better) | 2.58% | 2.63% | +0.05 points (−0.19 to +0.28) |
| Spoken questions, 100 VoiceBench OpenBookQA items | 62.0% | 61.0% | −1.0 points (−5.0 to +2.0) |
| Image questions, 102 MMStar items | 65.7% | 59.8% | −5.9 points (−11.8 to 0.0) |
| Video questions, 50 TempCompass clips | 68.0% | 68.0% | 0.0 points (−8.0 to +8.0) |
| Speech output, 40 sentences read back by Parakeet (word error rate) | 2.27% | 2.93% | +0.67 points (0.0 to +1.7) |
Differences are paired over the same items. Every difference is within its 95% range of zero; image questions come closest to a real drop. On text, this build called the right function with the right arguments on 90% of 40 simple and 70% of 40 live simple BFCL v3 requests (original Omni: 90% and 75%), and solved 86% of the first 50 GSM8K math problems (original: 76%).
Memory
The weights take 2.7 GB (2.5 GB for the model, 0.14 GB for the speech codec). The KV cache (bf16) grows by 12 KB per token in each request, plus a fixed 2.8 MB per request; audio, images and video enter the context as tokens. The longest context is 128,000 tokens, which needs 1.5 GB of KV cache per request, so a 64 GB Mac holds many requests at the full length.
Run it
Accept the terms at the top of this page, then log in with hf auth login (or set HF_TOKEN).
uv venv --python 3.12 && source .venv/bin/activate
uv pip install "mlx-vlm==0.7.12" "mlx-audio==0.6.4"
hf download balllvin/LFM3-Omni-2B-Abliterated-NVFP4-MLX --local-dir lfm3-omni
python lfm3-omni/examples/infer_mlx.py --model lfm3-omni --audio question.wav --image photo.jpg --out speech.wav
infer_mlx.py asks a text question, transcribes the audio file (16 kHz mono WAV), describes the photo and writes a spoken sentence to speech.wav, using the prompts from Liquid AI's LFM3-Omni-2B card. For text and images over the OpenAI chat API, run python -m mlx_vlm.server --model balllvin/LFM3-Omni-2B-Abliterated-NVFP4-MLX --port 8080.
How it was made
We added the difference between orcarouter/LFM3-1.7B-abliterated and LiquidAI/LFM3-1.7B to the Omni language core (task arithmetic), which moves the core by 1.0%; this is our bf16 LFM3-Omni-2B-abliterated. Then we rounded the core's 72 feed-forward weight matrices to NVFP4 and kept everything else, including the audio, vision and speech parts, as Liquid AI released them. We did no training on Omni.
Limits
- The language core is abliterated, so the model answers most harmful requests. Use it for research, with a person checking its output.
- Each number is one run on one setup.
Licence
LFM Open License v1.0, from the base models (LICENSE; NOTICE.md lists our changes). Commercial use is licensed only below 10 million US dollars of annual revenue. This is an independent build, not affiliated with Liquid AI or OrcaRouter.
Terms of use
Access to this repository requires agreeing to the terms in the access form on this page. In short: you alone are responsible and liable for your use of this model and for everything that comes from it, it is provided as is with no warranty, and its authors are not liable for any result of that use. Full text: TERMS.md.
To download after you agree, log in with the Hugging Face CLI (hf auth login) or set HF_TOKEN.