README.md

---
library_name: mlx
pipeline_tag: text-generation
license: mit
language:
- en
base_model: miromind-ai/MiroThinker-v1.0-8B
tags:
- agent
- open-source
- miromind
- deep-research
- mlx
---

# johnsmith968530/miromind-ai-MiroThinker-v1.0-8B-MLX-8bit

This model [johnsmith968530/miromind-ai-MiroThinker-v1.0-8B-MLX-8bit](https://huggingface.co/johnsmith968530/miromind-ai-MiroThinker-v1.0-8B-MLX-8bit) was
converted to MLX format from [miromind-ai/MiroThinker-v1.0-8B](https://huggingface.co/miromind-ai/MiroThinker-v1.0-8B)
using mlx-lm version **0.28.3**.

```bash
export MODEL1_MAJOR="miromind-ai"
export MODEL1_MINOR="MiroThinker-v1.0-8B"
export MODEL1_Q_BITS="8"
export MODEL1_DTS="$(date -z UTC +%Y%m%d_%H%M%SZ)"

echodo () { echo "$@" && "$@"; }
mkdir -p /tmp/mlx/lm/convert

# uv tool install -U mlx-lm

echodo time mlx_lm.convert \
  --hf-path "$MODEL1_MAJOR/$MODEL1_MINOR" \
  --mlx-path "/tmp/mlx/lm/convert/$MODEL1_DTS" \
  --quantize \
  --q-bits "$MODEL1_Q_BITS" \
  --upload-repo \
    "johnsmith968530/$MODEL1_MAJOR-$MODEL1_MINOR-MLX-${MODEL1_Q_BITS}bit"

# I tested the result using LM Studio 0.3.31
# on a MacBook Air, 13 inch, M4, 2025 with 32 GB unified memory
# macOS Tahoe 26.1.
```