matt macosko

divinetribe

AI & ML interests

local inference, apple silicon, mlx, quantization, llama, gemma, robotics, computer vision, yolov8, coreml, voice cloning, autonomous agents

Recent Activity

updated a Space about 20 hours ago

divinetribe/yolov8-oiv7-demo

updated a model about 20 hours ago

divinetribe/AesCode-32B-MLX-4bit

updated a model about 20 hours ago

divinetribe/AesCode-32B-MLX-8bit

View all activity

Organizations

New activity in lmstudio-community/Qwen3.8-27B-MLX-4bit 14 days ago

Tested on an M3 Max 36GB: the only model that still caught its own mistakes when a wrong answer would be expensive

#1 opened 2 months ago by

harryherbig

New activity in prism-ml/Ternary-Bonsai-27B-mlx-2bit 14 days ago

Ternary-Bonsai-27B-mlx-2bit slow on macbook 4 pro with oMLX

#2 opened 4 months ago by

nps798

New activity in manjunathshiva/gpt-oss-20b-tq3 14 days ago

Unified memory this needs on a Mac: 11.5 GB at 8k context

#2 opened 14 days ago by

divinetribe

New activity in deadbydawn101/RavenX-Trade-8B-MAP-128k-mlx-4bit 14 days ago

Unified memory this needs on a Mac: 18.8 GB at 8k context

#1 opened 14 days ago by

divinetribe

New activity in ddalcu/ACE-Step-1.5-XL-Turbo-MLX-Serve-8bit 14 days ago

Unified memory this needs on a Mac: 8.1 GB at 8k context

👍 1

#1 opened 14 days ago by

divinetribe

New activity in Jundot/DeepSeek-V4-Flash-0731-oQ2e-mtp 14 days ago

Unified memory this needs on a Mac: 107.2 GB at 8k context

#2 opened 14 days ago by

divinetribe

New activity in PocketAiHub/LFM2.5-8B-A1B-Abliterated 14 days ago

Unified memory this needs on a Mac: 7.1 GB at 8k context

#1 opened 14 days ago by

divinetribe

New activity in Youssofal/MiniMax-M2.7-Abliterated-Heretic-MLX-3bit 14 days ago

Unified memory this needs on a Mac: 108.6 GB at 8k context

#1 opened 14 days ago by

divinetribe

New activity in Jiunsong/SuperQwen-AgentWorld-35B-A3B-abliterated-mlx-4bit 14 days ago

Unified memory this needs on a Mac: 21.4 GB at 8k context

#1 opened 14 days ago by

divinetribe

New activity in peculiar-ragdoll/Sharp-MiniCPM5-2B-MLX-oQ6e 14 days ago

Unified memory this needs on a Mac: 3.6 GB at 8k context

#1 opened 14 days ago by

divinetribe

New activity in dealignai/MiniMax-M2.7-JANGTQ-CRACK 14 days ago

Unified memory this needs on a Mac: 64.0 GB at 8k context

#2 opened 14 days ago by

divinetribe

New activity in abenzerps/Spark-X2.5-4B-MLX-8bit 14 days ago

Unified memory this needs on a Mac: 6.8 GB at 8k context

🔥 1

#2 opened 14 days ago by

divinetribe

New activity in divinetribe/Nemotron-3-Nano-Omni-30B-Abliterated-MM-bf16 3 months ago

Considerably better than the other abliterated model flying around

#1 opened 3 months ago by

wacomctl672

New activity in mlx-community/Muse-Glimmer-30B-4bit 3 months ago

Can we get this with "tool use" available?

#1 opened 3 months ago by

kolyno25

New activity in Youssofal/Qwen3.6-27B-Abliterated-Heretic-Uncensored-MLX-4bit 3 months ago

Model fails to load in LM Studio MLX backend with shape mismatch error in vision_tower.patch_embed.proj.weight. Please check tensor layout.

#1 opened 4 months ago by

ohmygush

New activity in georgeis55/Muse-Glimmer-30B-MLX-6bit 3 months ago

LM Studio MLX doesn’t support "muse_glimmer"

👍 1

#1 opened 3 months ago by

atdaniel

New activity in google/gemma-4-12B-it 3 months ago

MLX 4-bit conversion of gemma-4-12B-it — with vision + audio embedder weights preserved

#27 opened 5 months ago by

jokernifty

New activity in meta-models/Muse-Glimmer-30B 3 months ago

How to get Muse-Glimmer working on a Mac / MacBook Pro in LM Studio and the M5 Metal Tensor issue

#42 opened 3 months ago by

Portll-LLM

New activity in Plachta/Seed-VC 3 months ago

Thank you — seed-vc is doing the voice conversion in a local music app

#2 opened 3 months ago by

divinetribe

New activity in NousResearch/Hermes-4-14B 3 months ago

generation_config.json eos_token_id is the pad token (151643), causing runaway generation

#6 opened 3 months ago by

divinetribe

matt macosko

AI & ML interests

Recent Activity

Organizations

divinetribe's activity

Tested on an M3 Max 36GB: the only model that still caught its own mistakes when a wrong answer would be expensive

Ternary-Bonsai-27B-mlx-2bit slow on macbook 4 pro with oMLX

Unified memory this needs on a Mac: 11.5 GB at 8k context

Unified memory this needs on a Mac: 18.8 GB at 8k context

Unified memory this needs on a Mac: 8.1 GB at 8k context

Unified memory this needs on a Mac: 107.2 GB at 8k context

Unified memory this needs on a Mac: 7.1 GB at 8k context

Unified memory this needs on a Mac: 108.6 GB at 8k context

Unified memory this needs on a Mac: 21.4 GB at 8k context

Unified memory this needs on a Mac: 3.6 GB at 8k context

Unified memory this needs on a Mac: 64.0 GB at 8k context

Unified memory this needs on a Mac: 6.8 GB at 8k context

Considerably better than the other abliterated model flying around

Can we get this with "tool use" available?

Model fails to load in LM Studio MLX backend with shape mismatch error in vision_tower.patch_embed.proj.weight. Please check tensor layout.

LM Studio MLX doesn’t support "muse_glimmer"

MLX 4-bit conversion of gemma-4-12B-it — with vision + audio embedder weights preserved

How to get Muse-Glimmer working on a Mac / MacBook Pro in LM Studio and the M5 Metal Tensor issue

Thank you — seed-vc is doing the voice conversion in a local music app

generation_config.json eos_token_id is the pad token (151643), causing runaway generation