mlx models.
sourced.
ready to run.

q4 · audio

MiniMax-Music3 (4-bit)

mlx-community

Text-to-music model from MiniMax that generates full songs.

9.2 GB 14 GB ram minimax-music3-community

◆ huggingface.co

details →

mixed · llm

MiniMax-M2.7 (4-bit)

mlx-community

Language model from MiniMax for agent workflows, coding and productivity tasks.

228.69B · 128.67 GB 157 GB ram modified-mit

◆ huggingface.co

details →

mixed · vlm

MiniMax-M3 (4-bit)

mlx-community

Mixture-of-experts multimodal model from MiniMax with text, image and video input.

427.04B · 241.48 GB 292 GB ram minimax-community

◆ huggingface.co

details →

bf16 · video

Lance 3B Video (bf16)

mlx-community

Text-to-video model from ByteDance.

15.62 GB 21 GB ram apache-2.0

◆ huggingface.co

details →

mixed · video

Bernini v2 (4-bit)

mlx-community

Video generation and editing model from ByteDance.

48.26 GB 60 GB ram apache-2.0

◆ huggingface.co

details →

q4 · video

LTX-2.5 (4-bit)

ddalcu

Video generation model from Lightricks with matching audio.

38.76 GB 49 GB ram ltx-2-community

◆ huggingface.co

details →

mixed · video

MiniMax-H3 (8-bit)

ddalcu

Video generation model from MiniMax with synchronized audio.

70.06 GB 87 GB ram minimax-h3-community

◆ huggingface.co

details →

q8 · image

ERNIE-Image (8-bit)

mflux-community

Text-to-image model from Baidu.

12.35 GB 17 GB ram apache-2.0

◆ huggingface.co

details →

bf16 · image

HiDream-O1 Image Dev (bf16)

mlx-community

Image generation and editing model from HiDream.ai, with reference-image input.

8.77B · 17.61 GB 24 GB ram mit

◆ huggingface.co

details →

q8 · image

Krea 2 Turbo (8-bit)

mflux-community

Text-to-image model from Krea.

22.17 GB 29 GB ram krea-2-community

◆ huggingface.co

details →

q4 · image

SenseNova U1.5 8B MoT (4-bit)

mlx-community

Unified model from SenseTime that generates, edits and understands images.

11.77 GB 17 GB ram apache-2.0

◆ huggingface.co

details →

q4 · image

Qwen-Image 2.1 (4-bit)

mlx-community

Image generation and editing model from Qwen.

9.59 GB 14 GB ram qwen-research

◆ huggingface.co

details →

q8 · image

Ming-Image 0.1 Design (8-bit)

joeynyc

Text-to-image model from inclusionAI for design work: posters, cards, UI and typography.

25.21 GB 33 GB ram mit

◆ huggingface.co

details →

q4 · llm

Kimi K2.5 (4-bit)

mlx-community

Mixture-of-experts language model from Moonshot AI, 1T parameters with 32B active.

1026.88B · 657.62 GB 792 GB ram modified-mit

◆ huggingface.co

details →

q4 · llm

Qwen3 8B (4-bit)

mlx-community

General-purpose chat model from Qwen.

8.19B · 4.61 GB 8 GB ram apache-2.0

◆ huggingface.co

details →

mixed · vlm

Gemma 4 E2B (4-bit)

lmstudio-community

Small chat model from Google with text, image and audio input.

5.12B · 4.34 GB 8 GB ram apache-2.0

◆ huggingface.co

details →

q8 · audio

Qwen3-ASR 0.6B (8-bit)

mlx-community

Speech recognition model from Qwen that transcribes audio to text.

0.94B · 1.01 GB 4 GB ram apache-2.0

◆ huggingface.co

details →

mixed · llm

gpt-oss 20B (MXFP4)

mlx-community

OpenAI's open-weight mixture-of-experts model for chat and reasoning.

20.91B · 12.08 GB 17 GB ram apache-2.0

◆ huggingface.co

details →

q4 · llm

Ornith 1.5 9B (4-bit)

ornith-ai

Dense 9B language model from ornith-ai for chat and text generation.

8.95B · 5.04 GB 9 GB ram

◆ huggingface.co

details →

mixed · vlm

Qwen3.6 35B A3B (4-bit)

lmstudio-community

Mixture-of-experts chat model from Qwen with text and image input.

35.95B · 20.4 GB 27 GB ram apache-2.0

◆ huggingface.co

details →

q4 · vlm

Qwen3.6 27B (4-bit)

lmstudio-community

General-purpose chat model from Qwen with text and image input.

27.78B · 16.05 GB 22 GB ram apache-2.0

◆ huggingface.co

details →

fp16 · audio

Whisper large-v3-turbo (fp16)

mlx-community

Speech recognition model from OpenAI that transcribes audio to text.

0.81B · 1.61 GB 4 GB ram mit

◆ huggingface.co

details →

q4 · vlm

Gemma 3 12B (4-bit)

mlx-community

Chat model from Google with text and image input.

12.19B · 8.03 GB 12 GB ram gemma

◆ huggingface.co

details →

q2 · vlm

Ternary Bonsai 27B (2-bit)

prism-ml

Ternary-weight compression of Qwen3.6 27B, for text generation with optional image input.

27.36B · 8.49 GB 13 GB ram apache-2.0

◆ huggingface.co

details →