mlx-optiq
← Models

Qwen3.5-9B-OptiQ-4bit

Qwen/Qwen3.5-9B, quantized with OptiQ by neuralforgequantum for Apple Silicon; takes images.

communityvisionMTP head
On disk8.2 GB
Memoryabout 11 GB
Capability Score69.85
Imagesyes

Capability Score breakdown

The six metrics behind this model's Capability Score of 69.85, as reported in the model card. Values are compared with uniform 4-bit.

  • This model
  • Uniform 4-bit
Capability Score breakdown: Qwen3.5-9B-OptiQ-4bitMMLU: This model 69.1, Uniform 4-bit 68.6; GSM8K: This model 81.7, Uniform 4-bit 81.7; IFEval: This model 71.2, Uniform 4-bit 71.0; BFCL-V3: This model 91.0, Uniform 4-bit 92.0; HumanEval: This model 81.1, Uniform 4-bit 78.7; HashHop: This model 25.0, Uniform 4-bit 26.0MMLU69.168.6GSM8K81.781.7IFEval71.271.0BFCL-V391.092.0HumanEval81.178.7HashHop25.026.0

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 neuralforgequantum/Qwen3.5-9B-OptiQ-4bit

OptiQ Code

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

optiq serve --model neuralforgequantum/Qwen3.5-9B-OptiQ-4bit
optiq code

OptiQ Lab

optiq lab --model neuralforgequantum/Qwen3.5-9B-OptiQ-4bit

Claude Code

optiq run claude -m neuralforgequantum/Qwen3.5-9B-OptiQ-4bit

Codex

optiq run codex -m neuralforgequantum/Qwen3.5-9B-OptiQ-4bit

OpenCode

optiq run opencode -m neuralforgequantum/Qwen3.5-9B-OptiQ-4bit

Mistral Vibe

optiq run vibe -m neuralforgequantum/Qwen3.5-9B-OptiQ-4bit

Hermes Agent

optiq run hermes -m neuralforgequantum/Qwen3.5-9B-OptiQ-4bit

OpenClaw

optiq run openclaw -m neuralforgequantum/Qwen3.5-9B-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("neuralforgequantum/Qwen3.5-9B-OptiQ-4bit")
print(generate(model, tokenizer, prompt="Hello", max_tokens=256))

Details

FamilyQwen · Qwen docs
ArchitectureQwen3.5
Base modelQwen/Qwen3.5-9B
Size on disk8.2 GB
Memory neededabout 11 GB
Imagesyes, through optiq serve and the Lab
MTP headincluded
Capability Score69.85 (from the model card)
Downloads63 in the last month
Updated2026-08-19
Hugging Faceneuralforgequantum/Qwen3.5-9B-OptiQ-4bit