mlx-optiq
← Models

LFM2.5-8B-A1B-OptiQ-4bit

LiquidAI/LFM2.5-8B-A1B, quantized by OptiQ for Apple Silicon; mixture-of-experts.

MoE
On disk5.5 GB
Memoryabout 8 GB
Capability Score32.11

Capability Score breakdown

The six metrics behind this model's Capability Score of 32.11, as reported in the model card.

Capability Score breakdown: LFM2.5-8B-A1B-OptiQ-4bitMMLU: This model 46.1; GSM8K: This model 54.6; IFEval: This model 32.9; BFCL-V3: This model 45.0; HumanEval: This model 14.0; HashHop: This model 0.0MMLU46.1GSM8K54.6IFEval32.9BFCL-V345.0HumanEval14.0HashHop0.0

Use it

Install

pip install mlx-optiq

Serve it

An OpenAI- and Anthropic-compatible server on this Mac. Every app below can use it.

optiq serve --model mlx-community/LFM2.5-8B-A1B-OptiQ-4bit

OptiQ Code

The terminal coding agent uses whatever model optiq serve is running.

optiq serve --model mlx-community/LFM2.5-8B-A1B-OptiQ-4bit
optiq code

OptiQ Lab

optiq lab --model mlx-community/LFM2.5-8B-A1B-OptiQ-4bit

Claude Code

optiq run claude -m mlx-community/LFM2.5-8B-A1B-OptiQ-4bit

Codex

optiq run codex -m mlx-community/LFM2.5-8B-A1B-OptiQ-4bit

OpenCode

optiq run opencode -m mlx-community/LFM2.5-8B-A1B-OptiQ-4bit

Mistral Vibe

optiq run vibe -m mlx-community/LFM2.5-8B-A1B-OptiQ-4bit

Hermes Agent

optiq run hermes -m mlx-community/LFM2.5-8B-A1B-OptiQ-4bit

OpenClaw

optiq run openclaw -m mlx-community/LFM2.5-8B-A1B-OptiQ-4bit

optiq run starts optiq serve if it is not running, sets the agent up for this session only, and stops the server when the agent exits.

mlx-lm, text only

from mlx_lm import load, generate

model, tokenizer = load("mlx-community/LFM2.5-8B-A1B-OptiQ-4bit")
print(generate(model, tokenizer, prompt="Hello", max_tokens=256))

Details

FamilyLFM2.5 · LFM2.5 docs
ArchitectureLFM2 MoE
Base modelLiquidAI/LFM2.5-8B-A1B
Size on disk5.5 GB
Memory neededabout 8 GB
Imagesno, text only
Capability Score32.11 (from the model card)
Downloads539 in the last month
Updated2026-10-14
Hugging Facemlx-community/LFM2.5-8B-A1B-OptiQ-4bit