Based on Qwen3.5-122B-A10B and converted from the MLX Community 4-bit release (https://huggingface.co/mlx-community/Qwen3.5-122B-A10B-4bit)

vision tools thinking

ollama run rekiniasty/qwen3.5:122b-a10b-mlx-int4

curl http://localhost:11434/api/chat \
  -d '{
    "model": "rekiniasty/qwen3.5:122b-a10b-mlx-int4",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='rekiniasty/qwen3.5:122b-a10b-mlx-int4',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'rekiniasty/qwen3.5:122b-a10b-mlx-int4',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Details

Updated 1 week ago

1 week ago

52996c6fd966 · 70GB ·

json

{ "architectures": [ "Qwen3_5MoeForConditionalGeneration" ], "eos_token_id": [ 248046, 248044 ], "im

4.1kB

json

{ "bos_token_id": 248044, "do_sample": true, "eos_token_id": [ 248046, 248044 ], "pad_token_id": 248

244B

json

{ "size": { "longest_edge": 16777216, "shortest_edge": 65536 }, "patch_size": 16, "temporal_patch_si

390B

json

{ "image_processor": { "data_format": "channels_first", "do_convert_rgb": true, "do_normalize": true

1.3kB

json

{ "version": "1.0", "truncation": null, "padding": null, "added_tokens": [ { "id": 248044, "content"

20MB

json

{ "add_prefix_space": false, "audio_bos_token": "<|audio_start|>", "audio_eos_token": "<|audio_end|>

1.1kB

json

{ "size": { "longest_edge": 25165824, "shortest_edge": 4096 }, "patch_size": 16, "temporal_patch_siz

385B

json

{ "!": 0, "\"": 1, "#": 2, "$": 3, "%": 4, "&": 5, "'": 6, "(": 7, ")": 8, "*": 9, "+": 10, ",": 11,

6.7MB

license

Apache License Version 2.0, January 2004 http://www.apache.org/licenses/ TERMS AND CONDITIONS FOR US

11kB

params

{ "min_p": 0, "presence_penalty": 1.5, "repeat_penalty": 1, "temperature": 1, "t

94B

template

13B

1212 tensors

70GB

Readme

Qwen3.5 122B A10B 4bit

Qwen3.5-122B-A10B quantized to 4-bit for efficient local deployment. Delivers strong reasoning, coding, and multilingual capabilities while significantly reducing memory requirements.

Source

Based on Qwen3.5-122B-A10B and converted from the MLX Community 4-bit release. https://huggingface.co/mlx-community/Qwen3.5-122B-A10B-4bit