Skip to content

What can the M5 Ultra 512 GB run?

What fits on the M5 Ultra 512 GB
Text models: 69Visual models: 0Strongest first
Memory
512 GB unified
Bandwidth
1,200 GB/s
Maker
Apple

Open the calculator on this machine

Text models that run on it

Text models that run on the M5 Ultra 512 GB, best first
ModelBest fitContextSpeedStrongest at
MiMo V2.6 Pro2-bit128K33 tok/sChat, Coding, Agents, Docs
DeepSeek V4.1 Flash3-bit1M62 tok/sChat, Coding, Agents, Docs
GLM-5.33-bit216K28 tok/sChat, Coding, Agents, Docs
GLM-5.3 FlashFP81M33 tok/sChat, Coding, Agents, Docs
Qwen3.8 Flash-Next8-bit256K66 tok/sChat, Coding, Agents
Muse Spark Open 816B A42B2-bit1M35 tok/sChat, Docs
MiMo V2.6 FlashFP4 + FP81M49 tok/sCoding, Agents
MiniMax M36-bit320K31 tok/sDocs
DeepSeek V4 FlashFP4 + FP81M33 tok/sDocs
DeepSeek V4 Pro1.6-bit1M47 tok/sChat, Coding, Agents, Docs
Kimi K2.62-bit80K42 tok/sChat, Coding, Agents, Docs
Hy38-bit224K25 tok/sChat, Coding, Agents, Docs
Nemotron 3 Ultra 550B A55B4-bit256K17 tok/sChat, Coding, Docs
Qwen3.5 397B A17B6-bit256K38 tok/sAgents, Docs
MiniMax M2.7FP8200K46 tok/sCoding, Agents
DeepSeek V3.23-bit160K32 tok/sChat, Coding, Agents, Docs
Mistral Medium 3.5 128BFP8256K5.7 tok/sChat, Coding, Agents, Docs
Qwen3.5 122B A10B8-bit256K49 tok/sChat, Coding, Agents, Docs
Nemotron 3 Super 120B A12B8-bit256K35 tok/sChat, Coding, Agents, Docs
Mistral Small 4 119BFP8256K74 tok/sChat, Docs
Inkling2-bit4K35 tok/sText and images
DeepSeek R13-bit160K30 tok/sNot rated
DeepSeek V33-bit160K30 tok/sNot rated
GLM 53-bit198K27 tok/sNot rated
GLM 4.5 Air8-bit128K37 tok/sNot rated
Inkling Small8-bit4K47 tok/sText and images
MiniMax M2.5FP8192K42 tok/sNot rated
GLM 5.13-bit198K27 tok/sNot rated
Laguna M.18-bit256K24 tok/sNot rated
Laguna S 2.18-bit1M55 tok/sNot rated
Qwen3 VL 235B A22B Instruct8-bit256K23 tok/sText and images
GLM 5.23-bit216K27 tok/sNot rated
Qwen3 235B A22B8-bit40K23 tok/sNot rated
Ornith 1.5 122B8-bit256K41 tok/sNot rated
Kimi K2 Instruct2-bit80K35 tok/sNot rated
Ornith 1.5 397B6-bit256K29 tok/sNot rated
Ornith 1.0 397B6-bit256K37 tok/sNot rated
Step 3.5 Flash8-bit256K52 tok/sNot rated
DeepSeek V3 03243-bit160K30 tok/sNot rated
DeepSeek R1 05283-bit160K30 tok/sNot rated
GLM 4.58-bit24K16 tok/sNot rated
Llama 4 Scout 17B 16E Instruct8-bit1.3M32 tok/sText and images
MiniMax M2FP8192K42 tok/sNot rated
MiMo V2.5FP81M34 tok/sNot rated
Solar Open 100B8-bit128K43 tok/sNot rated
GLM 5.3 Flash BF168-bit1M32 tok/sText and images
Qwen3 VL 235B A22B Thinking8-bit256K23 tok/sText and images
Qwen3 235B A22B Instruct 25078-bit256K23 tok/sNot rated
NVIDIA Nemotron 3.5 Super VL 120B A12B BF168-bit1M35 tok/sText and images
Hy3 Preview8-bit216K25 tok/sNot rated
Qwen3 Coder 480B A35B Instruct5-bit184K25 tok/sCoding
GLM 4.78-bit24K16 tok/sNot rated
GLM 5.3 BF163-bit216K27 tok/sNot rated
GLM 4.68-bit28K17 tok/sNot rated
DeepSeek V2 Chat8-bit160K27 tok/sNot rated
MiMo V2 FlashFP8256K36 tok/sNot rated
Le Triomphant ECE TW38-bit32K9.7 tok/sNot rated
TW3 JRGL V28-bit32K9.7 tok/sNot rated
Command A Plus 05 2026 Bf168-bit195K26 tok/sText and images
GLM 4.5V8-bit64K39 tok/sText and images

Best fit is the highest quality, up to 8-bit, that leaves the use case’s context. Ranked by our use-case scores, then by Hugging Face downloads. Speeds are rough, plus or minus 25%. The top 60 of 69 are shown.

Only just too big

These need up to a quarter more memory than the M5 Ultra 512 GB has. Spilling into system RAM or a lower precision can get them running, slowly.

Questions

What is the best LLM for the M5 Ultra 512 GB?

Of the open models we track, the strongest that run well on the M5 Ultra 512 GB are MiMo V2.6 Pro, DeepSeek V4.1 Flash, GLM-5.3. 69 text models fit it in all.

Calculated from published specs, not benchmarked. All machines