Qwen3.5-122B-A10B-OptiQ-2bit-mtp

Qwen/Qwen3.5-122B-A10B, quantized with OptiQ by scratchery for Apple Silicon; mixture-of-experts, takes images.

communityvisionMoE

On disk48.3 GB

Memoryabout 57 GB

Imagesyes

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 scratchery/Qwen3.5-122B-A10B-OptiQ-2bit-mtp

OptiQ Code

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

optiq serve --model scratchery/Qwen3.5-122B-A10B-OptiQ-2bit-mtp
optiq code

OptiQ Lab

optiq lab --model scratchery/Qwen3.5-122B-A10B-OptiQ-2bit-mtp

Claude Code

optiq run claude -m scratchery/Qwen3.5-122B-A10B-OptiQ-2bit-mtp

Codex

optiq run codex -m scratchery/Qwen3.5-122B-A10B-OptiQ-2bit-mtp

OpenCode

optiq run opencode -m scratchery/Qwen3.5-122B-A10B-OptiQ-2bit-mtp

Mistral Vibe

optiq run vibe -m scratchery/Qwen3.5-122B-A10B-OptiQ-2bit-mtp

Hermes Agent

optiq run hermes -m scratchery/Qwen3.5-122B-A10B-OptiQ-2bit-mtp

OpenClaw

optiq run openclaw -m scratchery/Qwen3.5-122B-A10B-OptiQ-2bit-mtp

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("scratchery/Qwen3.5-122B-A10B-OptiQ-2bit-mtp")
print(generate(model, tokenizer, prompt="Hello", max_tokens=256))

Details

Family	Qwen · Qwen docs
Architecture	Qwen3.5 MoE
Base model	Qwen/Qwen3.5-122B-A10B
Size on disk	48.3 GB
Memory needed	about 57 GB
Images	yes, through optiq serve and the Lab
Downloads	49 in the last month
Updated	2026-10-13
Hugging Face	scratchery/Qwen3.5-122B-A10B-OptiQ-2bit-mtp