Apps · Command line + local API

MLX LM

by Apple (ml-explore) · MIT · 7,475 GitHub stars

Apple's own command-line runner for MLX models: chat, generate, serve an OpenAI-style API, and fine-tune with LoRA on a Mac.

macOS

Get MLX LM

At a glance

Best for
The fastest generation on Apple silicon for small and mid-size models, when the whole model fits in memory.
Inference engine
  • MLX
GPU support
  • Apple MLX
Opens
  • MLX
Features
  • OpenAI-compatible API
  • Fine-tuning
Platforms
Apple silicon Macs only.
Speed
M4 Max, 4-bit: MLX decoded 21-87% faster than llama.cpp on models up to 8B (e.g. Qwen3-8B 93 vs 77 tok/s); the gap closes around 27B dense. llama.cpp still prefills long prompts faster.

Sources: blog.starmorph.com

Install

macOS

uv tool install mlx-lm

395 models on this site run in MLX LM

Based on the files published for each model: MLX. Open a model to see the exact command or file for your quantization.

Model Made by Type Size Opens it as Quality tier
gpt-oss-safeguard 120B OpenAI Decision 117B (5.1B active) MLX 80
gpt-oss-safeguard 20B OpenAI Decision 22B (3.6B active) MLX 76
Qwen3Guard Gen 0.6B Alibaba Decision 750M MLX 74
Prometheus 2 7B Prometheus Eval Decision 7.2B MLX 68
MiMo V2.6 Pro (1T-A42B) Xiaomi LLM 1,024B (42.0B active) MLX 99
GLM-5.3 (753B-A40B) Zhipu AI LLM 753B (40.0B active) MLX 97.8
MiMo V2.6 Flash (310B-A15B) Xiaomi LLM 311B (15.0B active) MLX 97.2
GLM-5.2 (753B-A40B) Zhipu AI LLM 753B (40.0B active) MLX 96.3
Hunyuan Hy4 Preview (780B) Tencent LLM 780B (55.0B active) MLX 96.3
Qwen3.8 Flash Next (125B-A6B) Alibaba LLM 180B (6.0B active) MLX 96.5
DeepSeek V4 Pro (April 2026) DeepSeek LLM 1,599B (49.0B active) MLX 95
GLM-5.1 (744B-A40B) Zhipu AI LLM 754B (40.0B active) MLX 95.2
GLM-5.3 Flash (320B-A18B) Zhipu AI LLM 321B (18.0B active) MLX 95.4
Qwen3.5 397B-A17B Alibaba LLM 403B (17.0B active) MLX 95
GLM-5 (744B-A40B) Zhipu AI LLM 754B (40.0B active) MLX 94.7
Kimi K2.5 (1T) Moonshot AI LLM 1,027B (32.0B active) MLX 94.3
Kimi K2.7 Code (1T) Moonshot AI LLM 1,027B (32.0B active) MLX 94.3
MiniMax M3 (427B) MiniMax LLM 427B (25.0B active) MLX 94.8
Qwen3.6 27B Alibaba LLM 28B MLX 94
DeepSeek R1 0528 (671B) DeepSeek LLM 685B (37.0B active) MLX 93
DeepSeek V3.2 (671B) DeepSeek LLM 685B (37.0B active) MLX 93.7
GLM-4.7 (355B-A32B) Zhipu AI LLM 358B (32.0B active) MLX 93.5
Kimi K2 Thinking (1T) Moonshot AI LLM 1,026B (32.0B active) MLX 93
MiniMax M2.7 (230B-A10B) MiniMax LLM 229B (10.0B active) MLX 93.5
DeepSeek R1 (671B) DeepSeek LLM 671B (37.0B active) MLX 92
DeepSeek V4 Flash (April 2026) DeepSeek LLM 291B (13.0B active) MLX 92
Gemma 4 26B-A4B Google LLM 26B (3.8B active) MLX 92.5
MiniMax M2.5 (230B-A10B) MiniMax LLM 229B (10.0B active) MLX 92.5
Qwen3.5 122B-A10B Alibaba LLM 125B (10.0B active) MLX 92
Step 3.5 Flash (196B-A11B) StepFun LLM 199B (11.0B active) MLX 92
DeepSeek V3.1 (671B) DeepSeek LLM 685B (37.0B active) MLX 91
DeepSeek V3.1 Terminus DeepSeek LLM 685B (37.0B active) MLX 91.5
Gemma 4 31B Google LLM 31B MLX 91
Kimi K2 Instruct 0905 (1T) Moonshot AI LLM 1,026B (32.0B active) MLX 91
MiniMax M2.1 (230B-A10B) MiniMax LLM 229B (10.0B active) MLX 91
Qwen3 235B-A22B Thinking 2507 Alibaba LLM 235B (22.0B active) MLX 91
Qwen3 Coder 480B-A35B Alibaba LLM 480B (35.0B active) MLX 91
MiMo V2 Flash (309B-A15B) Xiaomi LLM 310B (15.0B active) MLX 90.9
Qwen3 235B-A22B Instruct 2507 Alibaba LLM 235B (22.0B active) MLX 90
Kimi K2 Instruct (July 2025) Moonshot AI LLM 1,026B (32.0B active) MLX 89
MiniMax M2 (230B-A10B) MiniMax LLM 229B (10.0B active) MLX 89
Qwen3-VL 235B-A22B Instruct Alibaba LLM 236B (22.0B active) MLX 89
GLM-4.7 Flash (30B-A3B) Zhipu AI LLM 31B (3.0B active) MLX 88
K-EXAONE 236B-A23B LG AI Research LLM 237B (23.0B active) MLX 88
LongCat Flash Chat (560B) Meituan LLM 562B (27.0B active) MLX 88
Qwen3 235B-A22B Alibaba LLM 235B (22.0B active) MLX 88
Agents-A1 35B-A3B InternScience LLM 35B (3.0B active) MLX 87
Qwen AgentWorld 35B-A3B Alibaba LLM 35B (3.0B active) MLX 87
Devstral 2 (123B) Mistral AI LLM 125B MLX 86
Laguna S 2.1 (118B-A8B) Poolside LLM 118B (8.0B active) MLX 86
Qwen3 Coder Next (80B-A3B) Alibaba LLM 80B (3.0B active) MLX 86
Qwen3-Next 80B-A3B Thinking Alibaba LLM 81B (3.0B active) MLX 86
GLM-4.6 (355B-A32B) Zhipu AI LLM 357B (32.0B active) MLX 84
gpt-oss 120B OpenAI LLM 117B (5.1B active) MLX 84
Qwen3 Next 80B-A3B Instruct Alibaba LLM 81B (3.0B active) MLX 84
DeepSeek V3 0324 (671B) DeepSeek LLM 685B (37.0B active) MLX 83
Gemma 4 12B Google LLM 12B MLX 83
GLM-4.6V Zhipu AI LLM 108B (12.0B active) MLX 83
Command A Reasoning (111B) Cohere LLM 111B MLX 82
DeepSeek R1 Distill Llama 70B DeepSeek LLM 71B MLX 82
GLM-4.5 (355B-A32B) Zhipu AI LLM 358B (32.0B active) MLX 82
Leanstral 119B-A6B Mistral AI LLM 119B (6.5B active) MLX 82
Llama 3.3 70B Instruct Meta LLM 71B MLX 82
Command A (111B) Cohere LLM 111B MLX 80
ERNIE 4.5 300B-A47B Baidu LLM 300B (47.0B active) MLX 80
GLM-4.5V Zhipu AI LLM 108B (12.0B active) MLX 80
Hermes 4 405B Nous Research LLM 406B MLX 80
Kimi Dev 72B Moonshot AI LLM 73B MLX 80
Llama 3.3 Nemotron Super 49B v1.5 NVIDIA LLM 50B MLX 80
Mistral Medium 3.5 (128B) Mistral AI LLM 128B MLX 80.4
Mistral Small 4 (119B-A6B) Mistral AI LLM 119B (6.0B active) MLX 80.6
North Mini Code 1.0 (30B-A3B) Cohere LLM 30B (3.0B active) MLX 80
Seed-OSS 36B Instruct ByteDance LLM 36B MLX 80
Solar Open 100B Upstage LLM 103B (12.0B active) MLX 80
Trinity Large Thinking (400B-A13B) Arcee AI LLM 399B (13.0B active) MLX 80.3
DiffusionGemma 26B-A4B IT Google LLM 26B (4.0B active) MLX 78
GLM-4.5 Air (106B-A12B) Zhipu AI LLM 110B (12.0B active) MLX 78
Mistral Large 2 (123B) Mistral AI LLM 123B MLX 78
Qwen3 30B-A3B Thinking 2507 Alibaba LLM 31B (3.3B active) MLX 78
Qwen3 Coder 30B-A3B Alibaba LLM 31B (3.3B active) MLX 78
Qwen3-VL 32B Instruct Alibaba LLM 33B MLX 78
QwQ 32B Alibaba LLM 33B MLX 78
Qwen3 30B-A3B Instruct 2507 Alibaba LLM 31B (3.3B active) MLX 77
Devstral Small 2 (24B) Mistral AI LLM 24B MLX 76
Granite 4.2 30B IBM LLM 29B MLX 76
Olmo 3.1 32B Think Ai2 LLM 32B MLX 76
OpenReasoning Nemotron 32B NVIDIA LLM 33B MLX 76
Qwen3 32B Alibaba LLM 33B MLX 76
Qwen3-VL 30B-A3B Thinking Alibaba LLM 31B (3.3B active) MLX 76
Qwen3.5 9B Alibaba LLM 9.7B MLX 76
DeepSeek V3 (671B) DeepSeek LLM 671B (37.0B active) MLX 75
EXAONE Deep 32B LG AI Research LLM 32B MLX 74
GLM-4 32B 0414 Zhipu AI LLM 33B MLX 74
GLM-Z1 32B (0414) Zhipu AI LLM 33B MLX 74
Hermes 4 70B Nous Research LLM 71B MLX 74
Hunyuan A13B (80B-A13B) Tencent LLM 80B (13.0B active) MLX 74
InternVL3 78B Shanghai AI Lab LLM 78B MLX 74
Kimi Linear 48B-A3B Moonshot AI LLM 49B (3.0B active) MLX 74
Qwen3 30B-A3B Alibaba LLM 31B (3.3B active) MLX 74
Qwen3-VL 30B-A3B Instruct Alibaba LLM 31B (3.3B active) MLX 74
Qwen2.5 72B Instruct Alibaba LLM 73B MLX 73
Qwen2.5-VL 72B Instruct Alibaba LLM 73B MLX 73
Apriel 1.5 15B Thinker ServiceNow LLM 15B MLX 72
DeepSeek R1 Distill Qwen 32B DeepSeek LLM 33B MLX 72
DeepSeek V2.5 1210 DeepSeek LLM 236B (21.0B active) MLX 72
EXAONE 4.0 32B LG AI Research LLM 32B MLX 72
Gemma 3 27B IT Google LLM 27B MLX 72
Granite 4.1 30B IBM LLM 29B MLX 72
Llama 3.1 405B Instruct Meta LLM 406B MLX 72
Llama 4 Maverick (17B x 128E) Meta LLM 402B (17.0B active) MLX 72
Llama 4 Scout 17B-16E Meta LLM 109B (17.0B active) MLX 72
Magistral Small 1.2 (24B) Mistral AI LLM 24B MLX 72
Olmo 3.1 32B Instruct Ai2 LLM 32B MLX 72
Phi-4 Reasoning Plus (14B) Microsoft LLM 15B MLX 72
Qwen3.6 35B-A3B Alibaba LLM 36B (3.0B active) MLX 72
Tülu 3 70B Ai2 LLM 71B MLX 72
Nemotron Cascade 2 30B-A3B NVIDIA LLM 32B (3.5B active) MLX 71
Ornith 1.0 35B-A3B DeepReinforce LLM 36B (3.0B active) MLX 71
Ornith 1.5 35B-A3B DeepReinforce LLM 36B (3.0B active) MLX 71
Qwen3.5 35B-A3B Alibaba LLM 36B (3.0B active) MLX 71
Command R+ (08-2024) 104B Cohere LLM 104B MLX 70
Devstral Small 1.1 (24B) Mistral AI LLM 24B MLX 70
gpt-oss 20B OpenAI LLM 21B (3.6B active) MLX 70
InternVL3 38B Shanghai AI Lab LLM 38B MLX 70
Laguna XS 2.1 (33B-A3B) Poolside LLM 33B (3.0B active) MLX 70
Magistral Small 2507 Mistral AI LLM 24B MLX 70
MedGemma 27B Text IT Google LLM 27B MLX 70
OpenHands LM 32B All Hands AI LLM 33B MLX 70
Phi-4 Reasoning (14B) Microsoft LLM 15B MLX 70
Qwen2.5 32B Instruct Alibaba LLM 33B MLX 70
Qwen3-Omni 30B-A3B Instruct Alibaba LLM 35B (3.3B active) MLX 70
Qwen3.5 27B Alibaba LLM 28B MLX 70
Qwen3.8 27B Alibaba LLM 28B MLX 70
TMax 27B Ai2 LLM 27B MLX 70
Gemma 4 E4B Google LLM 8.0B MLX 68
Granite 4.2 8B IBM LLM 8.8B MLX 68
Hermes 3 (Llama 3.1 70B) Nous Research LLM 71B MLX 68
Magistral Small 2506 Mistral AI LLM 24B MLX 68
Mistral Small 3.2 24B Mistral AI LLM 24B MLX 68
QVQ 72B Preview Alibaba LLM 73B MLX 68
Qwen2.5-VL 32B Instruct Alibaba LLM 33B MLX 68
Qwen3 14B Alibaba LLM 15B MLX 68
Qwen2.5 Coder 32B Instruct Alibaba LLM 33B MLX 67
AceReason Nemotron 14B NVIDIA LLM 15B MLX 66
Devstral Small 2505 Mistral AI LLM 24B MLX 66
Hermes 4 14B Nous Research LLM 15B MLX 66
Llama 3.1 Nemotron 70B Instruct NVIDIA LLM 71B MLX 66
Llama 3.2 90B Vision Instruct Meta LLM 89B MLX 66
Ministral 3 14B Reasoning Mistral AI LLM 14B MLX 66
Mistral Small 3.1 24B Mistral AI LLM 24B MLX 66
Qwen2.5 Coder 14B Instruct Alibaba LLM 15B MLX 66
QwQ 32B Preview Alibaba LLM 33B MLX 66
Reka Flash 3.1 (21B) Reka AI LLM 21B MLX 66
Trinity Mini (26B-A3B) Arcee AI LLM 26B (3.0B active) MLX 66
Command R+ 104B (04-2024) Cohere LLM 104B MLX 64
DeepCoder 14B Preview Agentica LLM 15B MLX 64
ERNIE 4.5 21B-A3B Baidu LLM 22B (3.0B active) MLX 64
InternVL3 14B Shanghai AI Lab LLM 15B MLX 64
Ministral 3 14B Mistral AI LLM 14B MLX 64
Mistral Small 3 24B (2501) Mistral AI LLM 24B MLX 64
Phi-4 (14B) Microsoft LLM 15B MLX 64
Qwen2 72B Instruct Alibaba LLM 73B MLX 64
Qwen2-VL 72B Instruct Alibaba LLM 73B MLX 64
Codestral 22B Mistral AI LLM 22B MLX 62
Command R (08-2024) 32B Cohere LLM 32B MLX 62
DeepSeek R1 0528 Qwen3 8B DeepSeek LLM 8.2B MLX 62
DeepSeek R1 Distill Qwen 14B DeepSeek LLM 15B MLX 62
Dolphin Mistral 24B Venice Edition Cognitive Computations LLM 24B MLX 62
Gemma 2 27B IT Google LLM 27B MLX 62
Gemma 4 E2B Google LLM 5.1B MLX 62
Granite 4.0 H Small (32B-A9B) IBM LLM 32B (9.0B active) MLX 62
Qwen2.5 14B Instruct Alibaba LLM 15B MLX 62
Qwen3-VL 8B Instruct Alibaba LLM 8.8B MLX 62
Qwen3.5 4B Alibaba LLM 4.7B MLX 62
Aya Expanse 32B Cohere LLM 32B MLX 60
Aya Vision 32B Cohere LLM 33B MLX 60
Dolphin 3.0 R1 Mistral 24B Cognitive Computations LLM 24B MLX 60
Gemma 3 12B IT Google LLM 12B MLX 60
GLM-4.1V 9B Thinking Zhipu AI LLM 10B MLX 60
Granite 4.1 8B IBM LLM 8.8B MLX 60
Kimi-VL A3B Thinking 2506 Moonshot AI LLM 16B (3.0B active) MLX 60
Nemotron Nano 9B v2 NVIDIA LLM 8.9B MLX 60
OLMo 2 32B Instruct Ai2 LLM 32B MLX 60
Qwen2.5 14B Instruct 1M Alibaba LLM 15B MLX 60
Qwen3 8B Alibaba LLM 8.2B MLX 60
Fara 1.5 4B Microsoft LLM 4.5B MLX 58
GLM-Z1 9B (0414) Zhipu AI LLM 9.4B MLX 58
Llama 3 70B Instruct Meta LLM 71B MLX 58
MiMo-VL 7B RL Xiaomi LLM 8.3B MLX 58
Qwen2.5-Math 72B Instruct Alibaba LLM 73B MLX 58
Seed-Coder 8B Reasoning ByteDance LLM 8.3B MLX 58
EXAONE Deep 7.8B LG AI Research LLM 7.8B MLX 56
Falcon-H1R 7B TII LLM 7.6B MLX 56
InternVL3 8B Shanghai AI Lab LLM 7.9B MLX 56
LFM2 24B-A2B Liquid AI LLM 24B (2.3B active) MLX 56
Mage-VL 4.7B Microsoft LLM 4.7B MLX 56
Ministral 3 8B Mistral AI LLM 8.9B MLX 56
Phi-3.5 MoE (42B-A6.6B) Microsoft LLM 42B (6.6B active) MLX 56
Qwen3 4B Thinking 2507 Alibaba LLM 4.0B MLX 56
Seed-Coder 8B Instruct ByteDance LLM 8.3B MLX 56
Mixtral 8x22B Instruct Mistral AI LLM 141B (39.0B active) MLX 55
Qwen2.5 Coder 7B Instruct Alibaba LLM 7.6B MLX 55
Command R 35B (v01) Cohere LLM 35B MLX 54
MiniCPM4 8B OpenBMB LLM 8.2B MLX 54
Mistral Small 22B (2409) Mistral AI LLM 22B MLX 54
Olmo 3 7B Think Ai2 LLM 7.3B MLX 54
Qwen3 4B Instruct 2507 Alibaba LLM 4.0B MLX 54
Qwen3-VL 4B Instruct Alibaba LLM 4.4B MLX 54
DeepSeek Coder V2 Lite (16B) DeepSeek LLM 16B (2.4B active) MLX 52
Falcon 3 10B Instruct TII LLM 10B MLX 52
GLM-4 9B 0414 Zhipu AI LLM 9.4B MLX 52
Granite 4.2 3B IBM LLM 3.7B MLX 52
Hunyuan 7B Instruct Tencent LLM 7.5B MLX 52
LFM2.5 8B-A1B Liquid AI LLM 8.5B (1.5B active) MLX 52
Ling Mini 2.0 (16B-A1.4B) Ant Group (inclusionAI) LLM 16B (1.4B active) MLX 52
Mistral Nemo 12B Instruct Mistral AI LLM 12B MLX 52
Nous Hermes 2 Mixtral 8x7B DPO Nous Research LLM 47B (12.9B active) MLX 52
Qwen1.5 110B Chat Alibaba LLM 111B MLX 52
Qwen2.5 7B Instruct Alibaba LLM 7.6B MLX 52
Qwen2.5-VL 7B Instruct Alibaba LLM 8.3B MLX 52
Aya 23 35B Cohere LLM 35B MLX 50
DeepSeek R1 Distill Llama 8B DeepSeek LLM 8.0B MLX 50
Falcon-H1 7B Instruct TII LLM 7.6B MLX 50
Fara 7B Microsoft LLM 8.3B MLX 50
Gemma 2 9B IT Google LLM 9.2B MLX 50
InternLM3 8B Instruct Shanghai AI Lab LLM 8.8B MLX 50
Mixtral 8x7B Instruct Mistral AI LLM 47B (12.9B active) MLX 50
Olmo 3 7B Instruct Ai2 LLM 7.3B MLX 50
Qwen2 57B-A14B Instruct Alibaba LLM 57B (14.0B active) MLX 50
Qwen2.5 7B Instruct 1M Alibaba LLM 7.6B MLX 50
Qwen3 4B Alibaba LLM 4.0B MLX 50
Yi 1.5 34B Chat 01.AI LLM 34B MLX 50
GLM-4 9B Chat 1M Zhipu AI LLM 9.5B MLX 48
Hermes 3 (Llama 3.1 8B) Nous Research LLM 8.0B MLX 48
Llama 3.1 8B Instruct Meta LLM 8.0B MLX 48
DeepHermes 3 Llama 3 8B Preview Nous Research LLM 8.0B MLX 46
DeepSeek R1 Distill Qwen 7B DeepSeek LLM 7.6B MLX 46
Dolphin 3.0 Llama 3.1 8B Cognitive Computations LLM 8.0B MLX 46
EXAONE 3.5 7.8B LG AI Research LLM 7.8B MLX 46
Falcon 3 7B Instruct TII LLM 7.5B MLX 46
Gemma 3n E4B Google LLM 7.9B MLX 46
Granite 3.3 8B IBM LLM 8.2B MLX 46
Granite 4.1 3B IBM LLM 3.4B MLX 46
Ministral 8B (2410) Mistral AI LLM 8.0B MLX 46
Phi-3 Medium (14B) Microsoft LLM 14B MLX 46
Phi-3 Medium 128K Instruct Microsoft LLM 14B MLX 46
Phi-4 Mini Reasoning (3.8B) Microsoft LLM 3.8B MLX 46
Tülu 3 8B Ai2 LLM 8.0B MLX 46
Gemma 3 4B IT Google LLM 4.3B MLX 45
Ornith 1.0 9B DeepReinforce LLM 9.7B MLX 45
Ornith 1.5 9B DeepReinforce LLM 9.7B MLX 45
Aya Vision 8B Cohere LLM 8.6B MLX 44
Code Llama 70B Instruct Meta LLM 69B MLX 44
Command R7B Cohere LLM 8.0B MLX 44
Dolphin 2.9.4 Llama 3.1 8B Cognitive Computations LLM 8.0B MLX 44
Granite 4.0 H Tiny (7B-A1B) IBM LLM 6.9B (1.0B active) MLX 44
Hermes 2 Pro Llama 3 8B Nous Research LLM 8.0B MLX 44
Jamba Reasoning 3B AI21 Labs LLM 3.2B MLX 44
LFM2.5 2.6B Liquid AI LLM 2.7B MLX 44
LFM2.5-VL 3B Liquid AI LLM 3.1B MLX 44
Llama 3.1 Nemotron Nano 4B v1.1 NVIDIA LLM 4.5B MLX 44
MedGemma 4B IT Google LLM 4.3B MLX 44
Ministral 3 3B Mistral AI LLM 3.9B MLX 44
North Micro Vision Cohere LLM 2.5B MLX 44
OLMo 2 13B Instruct Ai2 LLM 14B MLX 44
OpenChat 3.6 8B OpenChat LLM 8.0B MLX 44
Phi-4 Mini (3.8B) Microsoft LLM 3.8B MLX 44
Qwen1.5 32B Chat Alibaba LLM 33B MLX 44
Qwen2 7B Instruct Alibaba LLM 7.6B MLX 44
Qwen2-VL 7B Instruct Alibaba LLM 8.3B MLX 44
TMax 9B Ai2 LLM 9.0B MLX 44
MiMo 7B RL Xiaomi LLM 7.8B MLX 43
Hunyuan 4B Instruct Tencent LLM 4.2B MLX 42
Llama 3 8B Instruct Meta LLM 8.0B MLX 42
Llama 3.2 11B Vision Instruct Meta LLM 11B MLX 42
Mistral-NeMo-Minitron 8B Instruct NVIDIA LLM 8.4B MLX 42
SmolLM3 3B Hugging Face LLM 3.1B MLX 42
Apertus 8B Instruct Swiss AI LLM 8.1B MLX 40
Aya Expanse 8B Cohere LLM 8.0B MLX 40
DeepSeek Prover V2 7B DeepSeek LLM 6.9B MLX 40
DeepSeek V2 Lite Chat DeepSeek LLM 16B (2.4B active) MLX 40
Granite 4.0 Micro (3B) IBM LLM 3.4B MLX 40
InternLM2.5 7B Chat Shanghai AI Lab LLM 7.7B MLX 40
Nous Hermes 2 SOLAR 10.7B Nous Research LLM 11B MLX 40
Phi-3.5 Mini Instruct Microsoft LLM 3.8B MLX 40
Qwen3.5 2B Alibaba LLM 2.3B MLX 40
SOLAR 10.7B Instruct Upstage LLM 11B MLX 40
StarCoder2 15B Instruct BigCode LLM 16B MLX 40
Unlimited OCR (3B-A0.6B) Baidu LLM 3.3B (0.6B active) MLX 40
Yi Coder 9B Chat 01.AI LLM 8.8B MLX 40
Code Llama 34B Instruct Meta LLM 34B MLX 38
Dolphin 2.9 Llama 3 8B Cognitive Computations LLM 8.0B MLX 38
Falcon3 Mamba 7B Instruct TII LLM 7.3B MLX 38
Granite 4.0 H Micro IBM LLM 3.2B MLX 38
LFM2 8B-A1B Liquid AI LLM 8.3B (1.5B active) MLX 38
MiniCPM3 4B OpenBMB LLM 4.1B MLX 38
Mistral 7B Instruct v0.3 Mistral AI LLM 7.3B MLX 38
Phi-3.5 Vision Instruct Microsoft LLM 4.2B MLX 38
Codestral Mamba 7B Mistral AI LLM 7.3B MLX 36
Falcon-H1 3B Instruct TII LLM 3.2B MLX 36
Gemma 3n E2B Google LLM 5.4B MLX 36
Hermes 2 Pro Mistral 7B Nous Research LLM 7.2B MLX 36
OLMo 2 7B Instruct Ai2 LLM 7.3B MLX 36
Phi-3 Mini 128K Instruct Microsoft LLM 3.8B MLX 36
Phi-3 Mini 4K Instruct Microsoft LLM 3.8B MLX 36
Qwen2.5 Coder 3B Instruct Alibaba LLM 3.1B MLX 36
Qwen2.5-Math 7B Instruct Alibaba LLM 7.6B MLX 36
Qwen3-VL 2B Instruct Alibaba LLM 2.1B MLX 36
Starling LM 7B Beta Nexusflow LLM 7.2B MLX 36
Voxtral Mini 3B Mistral AI LLM 4.7B MLX 36
Yi 1.5 9B Chat 01.AI LLM 8.8B MLX 36
Llama 3.2 3B Instruct Meta LLM 3.2B MLX 35
OpenChat 3.5 (0106) OpenChat LLM 7.2B MLX 35
Aya 23 8B Cohere LLM 8.0B MLX 34
CodeGemma 7B Google LLM 8.5B MLX 34
LFM2 2.6B Liquid AI LLM 2.6B MLX 34
Mathstral 7B Mistral AI LLM 7.3B MLX 34
Mistral 7B Instruct v0.2 Mistral AI LLM 7.2B MLX 34
Nous Hermes 2 Mistral 7B DPO Nous Research LLM 7.2B MLX 34
Phi-3 Vision 128K Instruct Microsoft LLM 4.2B MLX 34
Qwen1.5 14B Chat Alibaba LLM 14B MLX 34
Qwen2.5 3B Instruct Alibaba LLM 3.1B MLX 34
Qwen2.5-VL 3B Instruct Alibaba LLM 3.8B MLX 34
Hermes 3 Llama 3.2 3B Nous Research LLM 3.2B MLX 32
LFM2.5 1.2B Instruct Liquid AI LLM 1.2B MLX 32
OpenHermes 2.5 Mistral 7B Teknium LLM 7.2B MLX 32
StarCoder2 15B (base) BigCode LLM 16B MLX 32
Code Llama 13B Instruct Meta LLM 13B MLX 30
EXAONE 4.0 1.2B LG AI Research LLM 1.3B MLX 30
Falcon 3 3B Instruct TII LLM 3.2B MLX 30
Granite 3.3 2B IBM LLM 2.5B MLX 30
Mistral 7B Instruct v0.1 Mistral AI LLM 7.2B MLX 30
Qwen3 1.7B Alibaba LLM 2.0B MLX 30
Gemma 1.1 7B IT Google LLM 8.5B MLX 28
Gemma 2 2B Google LLM 2.6B MLX 28
Granite 4.0 H 1B IBM LLM 1.5B MLX 28
StableLM 2 12B Chat Stability AI LLM 12B MLX 28
Yi 1.5 6B Chat 01.AI LLM 6.1B MLX 28
DeepScaleR 1.5B Preview Agentica LLM 1.8B MLX 26
Falcon-H1 1.5B Instruct TII LLM 1.6B MLX 26
Granite Vision 3.2 2B IBM LLM 3.0B MLX 26
LFM2 1.2B Liquid AI LLM 1.2B MLX 26
OLMoE 1B-7B Instruct Ai2 LLM 6.9B (1.3B active) MLX 26
Qwen1.5 7B Chat Alibaba LLM 7.7B MLX 26
Qwen2.5 Coder 1.5B Instruct Alibaba LLM 1.5B MLX 26
Qwen3.5 0.8B Alibaba LLM 0.9B MLX 26
StarCoder2 3B BigCode LLM 3.0B MLX 26
Zephyr 7B Beta Hugging Face LLM 7.2B MLX 26
Code Llama 7B Instruct Meta LLM 6.7B MLX 24
LFM2-VL 1.6B Liquid AI LLM 1.6B MLX 24
Qwen1.5 MoE A2.7B Chat Alibaba LLM 14B (2.7B active) MLX 24
Qwen2-VL 2B Instruct Alibaba LLM 2.2B MLX 24
Qwen2.5 1.5B Instruct Alibaba LLM 1.5B MLX 24
SmolVLM2 2.2B Instruct Hugging Face LLM 2.3B MLX 24
StarCoder2 7B BigCode LLM 7.2B MLX 24
DeepSeek R1 Distill Qwen 1.5B DeepSeek LLM 1.8B MLX 22
Hunyuan 1.8B Instruct Tencent LLM 1.8B MLX 22
Llama 3.2 1B Instruct Meta LLM 1.2B MLX 22
Nemotron Mini 4B Instruct NVIDIA LLM 4.2B MLX 22
SmolVLM 2.2B Instruct Hugging Face LLM 2.3B MLX 22
Gemma 3 1B Google LLM 1.0B MLX 20
Qwen2.5-Math 1.5B Instruct Alibaba LLM 1.5B MLX 20
SmolLM2 1.7B Instruct Hugging Face LLM 1.7B MLX 20
Stable Code Instruct 3B Stability AI LLM 2.8B MLX 20
Yi 6B Chat 01.AI LLM 6.1B MLX 20
LFM2 700M Liquid AI LLM 0.7B MLX 18
Falcon 3 1B Instruct TII LLM 1.7B MLX 16
Phi-2 2.7B Microsoft LLM 2.8B MLX 16
Qwen2 1.5B Instruct Alibaba LLM 1.5B MLX 16
CodeGemma 2B Google LLM 2.5B MLX 14
Falcon-H1 0.5B Instruct TII LLM 0.5B MLX 14
LFM2.5 350M Liquid AI LLM 0.4B MLX 14
Qwen3 0.6B Alibaba LLM 0.8B MLX 14
SmolLM 1.7B Instruct Hugging Face LLM 1.7B MLX 14
DeepSeek Coder 1.3B Instruct DeepSeek LLM 1.4B MLX 12
Granite 4.0 H 350M IBM LLM 0.3B MLX 12
LFM2 350M Liquid AI LLM 0.4B MLX 12
LFM2-VL 450M Liquid AI LLM 0.5B MLX 12
LFM2.5 230M Liquid AI LLM 0.2B MLX 10
Qwen2.5 0.5B Instruct Alibaba LLM 0.5B MLX 10
SmolVLM 500M Instruct Hugging Face LLM 0.5B MLX 10
StableLM 2 1.6B Chat Stability AI LLM 1.6B MLX 10
StableLM 2 Zephyr 1.6B Stability AI LLM 1.6B MLX 10
Hunyuan 0.5B Instruct Tencent LLM 0.5B MLX 8
Qwen2.5 Coder 0.5B Instruct Alibaba LLM 0.5B MLX 8
SmolLM2 360M Instruct Hugging Face LLM 0.4B MLX 8
ERNIE 4.5 0.3B Baidu LLM 0.4B MLX 6
Gemma 3 270M Google LLM 0.3B MLX 6
Qwen2 0.5B Instruct Alibaba LLM 0.5B MLX 6
SmolVLM 256M Instruct Hugging Face LLM 0.3B MLX 6
TinyLlama 1.1B Chat TinyLlama LLM 1.1B MLX 6
SmolLM 360M Instruct Hugging Face LLM 0.4B MLX 5
SmolLM2 135M Instruct Hugging Face LLM 0.1B MLX 4
Falcon-H1 Tiny 90M Instruct TII LLM 0.1B MLX 3
SmolLM 135M Instruct Hugging Face LLM 0.1B MLX 2

Alternatives

LM Studio Ollama llama.cpp vLLM Jan Text Embeddings Inference SGLang GLiNER Lemonade Infinity Unsloth Desktop Sentence Transformers Docker Model Runner LocalAI Atomic Chat

Which app should I use?

Estimates come from published hardware specs and our open formulas, not from running the model on your machine. Real speed varies with runtime, drivers and settings. Numbers marked “Measured” come from a linked benchmark.