Hardware › Mac Studio M5 Max (40-core GPU) 48GB+
Mac Studio M5 Max (40-core GPU) 48GB+ for local AI: price in Australia & what it can run
price unavailableno AU retailer price found
48 GBunified (36.0 GB usable)
614 GB/smemory bandwidth
120 tok/smeasured, Llama 2 7B Q4_0
Apple shows 'From' for this chip but says it requires a 48GB memory upgrade at extra cost, so the exact 48GB price isn't on the page.
Benchmark
119.92 tok/s on Llama 2 7B Q4_0, tg128 (llama.cpp llama-bench), Metal backend. Source: ggml-org/llama.cpp/discussions/4167 scoreboard (checked 2026-10-19).
Specs: https://www.apple.com/au/mac-studio/specs/
AU prices found
Price unavailable: no matching listing at the retailers we can check.
Reference prices (not used for value rankings):
- Apple AU: A$5,199 (from price, memory upgrade extra). source, checked 20 Oct 2026, 7:44 am AEST
Community measured
No reader results yet. Be the first to submit one →
Which models fit? (8k context)
| Model | Quant | Needs / usable | Fit | Est. tok/s |
|---|---|---|---|---|
Q4_K_M | 464.4 / 36.0 GB | no | – | |
Q4_K_M | 499.4 / 36.0 GB | no | – | |
Q4_K_M | 20.0 / 36.0 GB | fits | 250 estimate | |
Q4_K_M | 348.5 / 36.0 GB | no | – | |
Q4_K_M | 15.7 / 36.0 GB | fits | 375 estimate | |
Q4_K_M | 45.9 / 36.0 GB | no | – | |
Q4_K_M | 22.5 / 36.0 GB | fits | 250 estimate | |
Q4_K_M | 305.3 / 36.0 GB | no | – | |
MiniCPM V 4.7 1B new | Q4_K_M | 2.0 / 36.0 GB | fits | 484 estimate |
Q4_K_M | 464.4 / 36.0 GB | no | – | |
Q4_K_M | 195.9 / 36.0 GB | no | – | |
GLM 5.3 (MoE) new | Q4_K_M | 458.4 / 36.0 GB | no | – |
Q4_K_M | 110.3 / 36.0 GB | no | – | |
Q4_K_M | 18.4 / 36.0 GB | fits | 27 estimate | |
Q4_K_M | 20.2 / 36.0 GB | fits | 250 estimate | |
Q4_K_M | 9.1 / 36.0 GB | fits | 63 estimate | |
Q4_K_M | 23.0 / 36.0 GB | fits | 250 estimate | |
Q4_K_M | 81.4 / 36.0 GB | no | – | |
Q4_K_M | 22.1 / 36.0 GB | fits | 24 estimate | |
Q4_K_M | 17.2 / 36.0 GB | fits | 197 estimate | |
Q4_K_M | 6.0 / 36.0 GB | fits | 94 estimate | |
Q4_K_M | 7.1 / 36.0 GB | fits | 78 estimate | |
Q4_K_M | 4.1 / 36.0 GB | fits | 161 estimate | |
Q4_K_M | 77.0 / 36.0 GB | no | – | |
Q4_K_M | 20.4 / 36.0 GB | fits | 250 estimate | |
gpt-oss 20B (MoE) older | MXFP4 | 12.4 / 36.0 GB | fits | 238 estimate |
gpt-oss 120B (MoE) older | MXFP4 | 63.5 / 36.0 GB | no | – |
Qwen 3 4B older | Q4_K_M | 4.6 / 36.0 GB | fits | 187 estimate |
Qwen 3 8B older | Q4_K_M | 7.2 / 36.0 GB | fits | 92 estimate |
Qwen 3 32B older | Q4_K_M | 23.0 / 36.0 GB | fits | 23 estimate |
Llama 3.1 8B Instruct older | Q4_K_M | 6.9 / 36.0 GB | fits | 93 estimate |
Mistral 7B Instruct v0.3 older | Q4_K_M | 6.5 / 36.0 GB | fits | 103 estimate |