mlx-optiq
← Models

Qwen3.8-27B-OBLITERATED-OptiQ-4bit

OBLITERATUS/Qwen3.8-27B-OBLITERATED, quantized with OptiQ by npario for Apple Silicon; takes images.

communityvisionMTP head
On disk20.7 GB
Memoryabout 25 GB
Parent score87.98 Qwen3.8-27B-OptiQ-4bit
Imagesyes

This model reuses the per-layer bit allocation of Qwen3.8-27B-OptiQ-4bit, whose measured Capability Score is 87.98.

The parent's Capability Score breakdown

This model has no Capability Score of its own. The chart shows the six metrics behind its parent Qwen3.8-27B-OptiQ-4bit's score of 87.98, taken from the parent's model card.

The parent's Capability Score breakdown: Qwen3.8-27B-OptiQ-4bitMMLU: Qwen3.8-27B-OptiQ-4bit 84.1; GSM8K: Qwen3.8-27B-OptiQ-4bit 96.1; IFEval: Qwen3.8-27B-OptiQ-4bit 77.1; BFCL-V3: Qwen3.8-27B-OptiQ-4bit 93.5; HumanEval: Qwen3.8-27B-OptiQ-4bit 92.1; HashHop: Qwen3.8-27B-OptiQ-4bit 85.0MMLU84.1GSM8K96.1IFEval77.1BFCL-V393.5HumanEval92.1HashHop85.0

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 npario/Qwen3.8-27B-OBLITERATED-OptiQ-4bit

OptiQ Code

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

optiq serve --model npario/Qwen3.8-27B-OBLITERATED-OptiQ-4bit
optiq code

OptiQ Lab

optiq lab --model npario/Qwen3.8-27B-OBLITERATED-OptiQ-4bit

Claude Code

optiq run claude -m npario/Qwen3.8-27B-OBLITERATED-OptiQ-4bit

Codex

optiq run codex -m npario/Qwen3.8-27B-OBLITERATED-OptiQ-4bit

OpenCode

optiq run opencode -m npario/Qwen3.8-27B-OBLITERATED-OptiQ-4bit

Mistral Vibe

optiq run vibe -m npario/Qwen3.8-27B-OBLITERATED-OptiQ-4bit

Hermes Agent

optiq run hermes -m npario/Qwen3.8-27B-OBLITERATED-OptiQ-4bit

OpenClaw

optiq run openclaw -m npario/Qwen3.8-27B-OBLITERATED-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("npario/Qwen3.8-27B-OBLITERATED-OptiQ-4bit")
print(generate(model, tokenizer, prompt="Hello", max_tokens=256))

Details

FamilyQwen · Qwen docs
ArchitectureQwen3.5
Base modelOBLITERATUS/Qwen3.8-27B-OBLITERATED
Size on disk20.7 GB
Memory neededabout 25 GB
Imagesyes, through optiq serve and the Lab
MTP headincluded
Family referenceThis model reuses the per-layer bit allocation of Qwen3.8-27B-OptiQ-4bit, whose measured Capability Score is 87.98.
Downloads163 in the last month
Updated2026-09-04
Hugging Facenpario/Qwen3.8-27B-OBLITERATED-OptiQ-4bit