mlx-community/LFM3-1.7B-OptiQ-4bit

Built with mlx-optiq, the MLX-native toolkit to quantize, fine-tune, and serve LLMs locally on Apple Silicon, no PyTorch and no cloud. All OptiQ quants · Docs

An OptiQ mixed-precision quant of LiquidAI/LFM3-1.7B, the first model in Liquid's LFM3 family. It is 1.3 GB on disk, down from 3.5 GB at bf16, with a 256k context.

LFM3 keeps Liquid's hybrid backbone of short convolutions and attention, and adds sliding-window attention, so only 6 of its 24 layers keep a full-length KV cache. OptiQ measures each layer's sensitivity and assigns its bit-width from that.

What it is

Property	Value
Base	LiquidAI/LFM3-1.7B
Architecture	lfm3: 24 layers, 13 convolution, 5 sliding-window attention (256 tokens), 6 full attention
Method	OptiQ mixed-precision, sensitivity-driven (bf16 reference)
On disk	1.3 GB (bf16: 3.5 GB)
Context	256k

About half of the weight matrices stay at 8-bit (75 of 153), the ones whose quantization moved the output most. The rest are at 4-bit.

Capability Score

Six-metric mean, the standard OptiQ eval, run in reasoning mode because the model thinks before it answers by default. The baseline is a uniform 4-bit quant of the same model, scored the same way.

Metric	OptiQ	Uniform 4-bit
MMLU (generative)	77.6%	76.3%
GSM8K	92.4%	92.4%
IFEval (strict)	72.6%	66.5%
BFCL-V3 simple (200 calls)	91.5%	93.0%
HumanEval (164 problems, pass@1)	72.6%	64.0%
HashHop (long-context retrieval)	21.0%	14.0%
Capability Score (mean of 6)	71.29	67.70

The OptiQ quant is ahead by 3.6 points overall. The gains are on instruction following (+6.1) and code (+8.6). BFCL is 1.5 points behind, inside its ±3.5 point margin on 200 calls.

Run it

pip install -U mlx-optiq
optiq serve --model mlx-community/LFM3-1.7B-OptiQ-4bit

That gives you an OpenAI and Anthropic compatible endpoint with mixed-precision KV cache (kv_config.json in this repo), tool calling, and prompt caching.

The model reasons before it answers, so give it a generous max_tokens. Liquid's recommended sampling (temperature 0.1, top_k 50, repetition_penalty 1.1) ships in generation_config.json, and optiq serve applies it automatically.

Or from Python. mlx-lm does not know the LFM3 architecture yet, so register OptiQ's implementation first:

from optiq.models import install_vendored
from mlx_lm import load, generate

install_vendored()
model, tokenizer = load("mlx-community/LFM3-1.7B-OptiQ-4bit")
prompt = tokenizer.apply_chat_template(
    [{"role": "user", "content": "Summarise this contract clause in two sentences."}],
    add_generation_prompt=True, tokenize=False,
)
print(generate(model, tokenizer, prompt=prompt, max_tokens=4096))

Links

• Project website: mlx-optiq.com
• All OptiQ quants: mlx-optiq.com/models
• PyPI: pypi.org/project/mlx-optiq
• Base model: LiquidAI/LFM3-1.7B