mimo-v2.6:9b-mlx-bf16

641 3 hours ago

MiMo-V2.6-Distill-Qwen-9B is Xiaomi MiMo’s 9B model for code and agent workflows

vision tools thinking 9b 
ollama run mimo-v2.6:9b-mlx-bf16

Details

3 hours ago

98bfe7751130 · 19GB

json
{ "architectures": [ "Qwen3_5ForConditionalGeneration" ], "image_token_id": 248056, "model_type": "q
json
{ "eos_token_id": [ 248046, 248044 ], "do_sample": true, "temperature": 0.6, "top_k": 20, "top_p": 0
json
{ "do_convert_rgb": true, "do_normalize": true, "do_rescale": true, "do_resize": true, "image_mean":
json
{ "image_processor": { "do_convert_rgb": true, "do_normalize": true, "do_rescale": true, "do_resize"
json
{ "version": "1.0", "truncation": null, "padding": null, "added_tokens": [ { "id": 248044, "content"
json
{ "add_prefix_space": false, "audio_bos_token": "<|audio_start|>", "audio_eos_token": "<|audio_end|>
json
{ "size": { "longest_edge": 25165824, "shortest_edge": 4096 }, "patch_size": 16, "temporal_patch_siz
json
{ "!": 0, "\"": 1, "#": 2, "$": 3, "%": 4, "&": 5, "'": 6, "(": 7, ")": 8, "*": 9, "+": 10, ",": 11,
license
MIT (declared by XiaomiMiMo in the pinned source model card) https://huggingface.co/XiaomiMiMo/MiMo-
params
{ "min_p": 0, "presence_penalty": 0, "repeat_penalty": 1, "temperature": 0.6, "t
760 tensors

Readme

MiMo-V2.6-Distill-Qwen-9B

MiMo-V2.6-Distill-Qwen-9B is Xiaomi MiMo’s 9B model for code and agent workflows. Fine-tuned from Qwen3.5-9B on MiMo-generated data, it works with text and images and supports thinking and tool use.

What you can use it for

  • Software development: understand unfamiliar code, investigate failures, and draft fixes.
  • Agent workflows: work through multi-step tasks when connected to tools.
  • Visual coding: use screenshots or interface references to guide an implementation.
  • Security analysis: examine code and explain potential vulnerabilities.

With support for up to 256K tokens of context, the model can work across long files, documentation, and extended task histories.

Get started

ollama run mimo-v2.6

Evaluation

Xiaomi reports these results for MiMo-V2.6-Distill-Qwen-9B:

Benchmark Metric Qwen3.5-9B MiMo-V2.6-Distill-Qwen-9B
SWE Verified avg@3 60.0 61.1
SWE Pro avg@3 32.0 44.6
AutomationBench v1.0.6 avg@1 5.0 30.3
Terminal Bench 2.1 avg@1 27.0 37.1
Toolathlon-Verified avg@1 25.9 35.2

See Xiaomi’s model card and the MiMo-V2.6 technical report for the evaluation setup and more results. The model is released under the MIT license.