Apple · M5 Max · Desktop
Best AI Models for Mac Studio M5 Max (48 GB)
Runs 27–35B models well; 70B only just fits in 44.5 GB, and 120B MoE doesn't fit.
| # | Model | Speed | Score |
|---|---|---|---|
| 1 | Qwen3.8 27BAlibabaHow to run → Q4_K_M · 17.9 GB | 24.7 tok/s | |
| 2 | Qwen3.6 27BAlibaba Q4_K_M · 17.9 GB | 24.7 tok/s | |
| 3 | Gemma 4 31B ITGoogle Q4_K_M · 21.7 GB | 21.1 tok/s | |
| 4 | Gemma 4 26B A4B ITGoogle Q4_K_M · 16.5 GB | 80.7 tok/s | |
| 5 | Q4_K_M · 19.3 GB | 72.7 tok/s |
| # | Model | Speed | Score |
|---|---|---|---|
| 1 | Qwen3.8 27BAlibabaHow to run → Q4_K_M · 17.9 GB | 24.7 tok/s | |
| 2 | Gemma 4 31B ITGoogle Q4_K_M · 21.7 GB | 21.1 tok/s | |
| 3 | Qwen3.6 35B A3BAlibaba Q4_K_M · 22.0 GB | 67.8 tok/s | |
| 4 | Gemma 4 26B A4B ITGoogle Q4_K_M · 16.5 GB | 80.7 tok/s | |
| 5 | Q4_K_M · 19.3 GB | 72.7 tok/s |
| # | Model | Speed | Score |
|---|---|---|---|
| 1 | Qwen3.8 27BAlibabaHow to run → Q4_K_M · 17.9 GB | 24.7 tok/s | |
| 2 | Qwen3.5 27BAlibaba Q4_K_M · 17.9 GB | 24.7 tok/s | |
| 3 | Gemma 4 31B ITGoogle Q4_K_M · 21.7 GB | 21.1 tok/s | |
| 4 | Gemma 4 26B A4B ITGoogle Q4_K_M · 16.5 GB | 80.7 tok/s | |
| 5 | Q4_K_M · 19.3 GB | 72.7 tok/s |
| # | Model | Speed | Score |
|---|---|---|---|
| 1 | Qwen3.8 27BAlibabaHow to run → Q4_K_M · 17.9 GB | 24.7 tok/s | |
| 2 | Qwen3.6 27BAlibaba Q4_K_M · 17.9 GB | 24.7 tok/s | |
| 3 | Qwen3.6 35B A3BAlibaba Q4_K_M · 22.0 GB | 67.8 tok/s | |
| 4 | Gemma 4 31B ITGoogle Q4_K_M · 21.7 GB | 21.1 tok/s | overall |
| 5 | Q4_K_M · 19.3 GB | 72.7 tok/s |
| # | Model | Speed ↓ | Score |
|---|---|---|---|
| 1 | GPT OSS 20BOpenAIHow to run → Q4_K_M · 13.1 GB | 92.4 tok/s | |
| 2 | Gemma 4 26B A4B ITGoogle Q4_K_M · 16.5 GB | 80.7 tok/s | |
| 3 | Q4_K_M · 19.3 GB | 72.7 tok/s | |
| 4 | Qwen3 30B A3B Thinking 2507Alibaba Q4_K_M · 19.4 GB | 69.5 tok/s | |
| 5 | Qwen3 30B A3B Instruct 2507Alibaba Q4_K_M · 19.4 GB | 69.5 tok/s |
The Mac Studio M5 Max is the fastest 48 GB machine: it runs 27–35B models well. macOS keeps about 3.5 GB, leaving 44.5 GB for models. With 68 TFLOPS it reads long prompts faster than most 48 GB machines: a 4K-token prompt takes 17 s. M5 Max generation, released 2026.
Mac Studio M5 Max strengths and weaknesses
Strengths
- Runs 35B MoE models: Qwen3.6 35B A3B at 67.8 tok/s in 22 GB
- Room for long context: Qwen3.8 27B at Q4 uses 17.9 GB, leaving 26.6 GB free
Weaknesses
- 120B MoE models don't fit: GPT OSS 120B needs 70.8 GB at Q4 vs 44.5 GB usable here
- Slow prompt reading: A 32K-token prompt on Qwen3.8 27B waits 2.2 min for the first token
- No CUDA: vLLM, AWQ, GPTQ and FP8 builds don't run; use GGUF or MLX
Memory for models
44.5 GB of 48 GB120B MoE models need 70.8 GB at Q4, more than the 44.5 GB available.
Other memory sizes of the same machine first, then machines and GPUs people weigh against it, each with the one reason it is worth a look. Speeds on Qwen3.6 35B A3B.
Output speed depends on memory bandwidth (614 GB/s). Grey rows need more than the 44.5 GB available.
- 9BQwen3.5 9B · Q4_K_M · 6.6 GB67.6tok/s
- 14BQwen3 14B · Q4_K_M · 11.9 GB45.2tok/s
- 27–31BQwen3.8 27B · Q4_K_M · 17.9 GB24.7tok/s
- 35B MoEQwen3.6 35B A3B · Q4_K_M · 22.0 GB67.8tok/s
- 10.8tok/s
- 120B MoEGPT OSS 120B · does not fitneeds 70.8 GB
Estimated speeds: Q4_K_M, 4K context, one user. How we estimate
Max generations: writing vs compute
M1 Max → M5 MaxWriting speed follows memory bandwidth and barely moved: 54.5 to 67.8 tok/s since the M1 Max. Compute, which sets how fast prompts are read, grew 4.4×. Indexed to the M1 Max; speeds on Qwen3.6 35B A3B, same 32 GB.
All 1152 models on the Mac Studio M5 Max
Featured onlyBest quant that fits in 44.5 GB, ordered by llmrun Score. Fit grades how well it runs here, S best. How ranking works
| Model | Quant | VRAM | Speed | Context | Score | Fit |
|---|---|---|---|---|---|---|
Llama 3.2 3BChat | Q4_K_M | 4.1 GB | 202.7 t/s | — | S100, Runs great | |
Llama Guard 3 1BChat | Q4_K_M | 1.7 GB | 338.4 t/s | 131K | S100, Runs great | |
Ling 3.0 Flash VLVisionFunctions | IQ2_M | 43.2 GB | 75.8 t/s | 131K | B60, Decent | |
| Q4_K_M | 7.3 GB | 79.7 t/s | 131K | S88, Runs great | ||
Gemma 3 12B PtVision | Q4_K_M | 10.9 GB | 53.5 t/s | — | A81, Runs well | |
Mistral Small 4 119B 2603ChatFunctions | IQ2_XXS | 42.8 GB | 83.9 t/s | 1049K | B62, Decent | |
| Q4_K_M | 7.3 GB | 80.0 t/s | 131K | S88, Runs great | ||
| Q4_K_M | 11.8 GB | 45.2 t/s | 41K | A77, Runs well | ||
Orca Mini 3BChatMath | Q4_K_M | 2.3 GB | 190.2 t/s | 2K | S99, Runs great | |
Kimi Dev 72BChatCodeFunctions | Q3_K_L | 42.9 GB | 11.2 t/s | 131K | D25, Barely runs | |
LFM2 1.2B RAGChat | Q4_K_M | 1.2 GB | 401.7 t/s | 128K | S100, Runs great | |
| IQ4_XS | 43.6 GB | 11.1 t/s | 131K | D21, Barely runs | ||
| Q4_K_M | 1.2 GB | 401.7 t/s | 128K | S100, Runs great | ||
| Q3_K_M | 39.8 GB | 12.2 t/s | 66K | C38, Usable | ||
Qwen2.5 Coder 3B InstructChatCodeFunctions | Q4_K_M | 2.8 GB | 192.7 t/s | 33K | S99, Runs great |
Specs
- Chip
- M5 Max
- CPU cores
- 18
- GPU cores
- 40
- Unified memory
- 48 GB
- For models
- 44.5 GB (macOS keeps 3.5 GB)
- Memory bandwidth
- 614 GB/s
- FP16 (dense)
- 68 TFLOPS
- Neural Engine
- 38 TOPS
- Released
- 2026-09-22
Mac Studio M5 Max FAQ
How much memory is available for AI on the Mac Studio M5 Max (48 GB)?
About 44.5 GB of the 48 GB. macOS keeps roughly 3.5 GB for itself and other apps.
What is the largest model the Mac Studio M5 Max (48 GB) can run?
Qwen3.6 35B A3B (36.0B) at Q4_K_M uses 22.0 GB, 50% of the available memory, and runs at about 67.8 tok/s.
Can the Mac Studio M5 Max (48 GB) run Qwen3.8 27B?
Yes. Qwen3.8 27B fits at Q4_K_M, using 17.9 GB of the 44.5 GB available, and it is the best model that runs well on this machine.
Can the Mac Studio M5 Max (48 GB) run 120B MoE models?
No. GPT OSS 120B needs 70.8 GB at Q4, more than the 44.5 GB available.
How long does the Mac Studio M5 Max (48 GB) take to read a long prompt?
On Qwen3.8 27B, about 17 s for a 4K-token prompt and 2.2 min for a 32K-token document before the first word appears.
Is renting a GPU cheaper than buying the Mac Studio M5 Max?
An A40 (48 GB) rents from $0.35/hr.