Hardware field guide

Mac Studio M5 Max (32-core GPU) 36GB

Name: Mac Studio M5 Max (32-core GPU) 36GB local model estimates
Creator: Web Asset House

36 GiB listed memory; 27.0 GiB default usable budget. Unified memory is shared with the system.

Listed memory capacity36 GiB

Listed bandwidth460 GB/s

Power referenceUnknown

Approximate USD reference$2,499

Apple US advertised system price; 512GB SSD. Apple US public Product/Offer JSON-LD read on 2026-10-22. Exact configuration advertised price before tax; a point quote, not a street-price range. Delivery availability was not tested. SKU MHL64LL/A. Price source

Affiliate linking is inactive unless an enrolled publisher supplies a tag and verified product link. Required notice for an enrolled release: “As an Amazon Associate, Web Asset House earns from qualifying purchases.”

Check a model ↗View maker product page ↗

Current Apple specification. GPU core variant is material to memory bandwidth.
Display, keyboard and pointing device are separate.

Models to investigate by task.

Capacity-qualified choices, ordered by recorded downloads. This is a shortlist for testing, not a quality ranking.

Chat and general use

Qwen3-0.6B · at least one recorded quantization fits the estimated budget
Qwen3-4B · at least one recorded quantization fits the estimated budget
Qwen3-14B · at least one recorded quantization fits the estimated budget
Qwen3-8B · at least one recorded quantization fits the estimated budget
Qwen2.5-7B-Instruct · at least one recorded quantization fits the estimated budget
Qwen2.5-0.5B-Instruct · at least one recorded quantization fits the estimated budget

Code-focused models

Qwen2.5-Coder-14B-Instruct · at least one recorded quantization fits the estimated budget
Qwen2.5-Coder-7B-Instruct · at least one recorded quantization fits the estimated budget
Qwen3-Coder-30B-A3B-Instruct · at least one recorded quantization fits the estimated budget
Qwen2.5-Coder-32B-Instruct · at least one recorded quantization fits the estimated budget
Qwen2.5-Coder-3B-Instruct · at least one recorded quantization fits the estimated budget
Qwen2.5-Coder-1.5B-Instruct · at least one recorded quantization fits the estimated budget

Vision models

No supported memory-fit recommendation in this snapshot. Additional architecture or vision-memory evidence may be needed.

Model compatibility at a glance.

Q4_K_M where recorded, otherwise the first available format; up to 4,096 tokens
Model / quantization	Required memory	Fit estimate	Estimated generation
Qwen3-0.6BQ4_K_M	1.81 GiB	fits	463.8–927.6 estimated tokens/s
Qwen3-4BQ4_K_M	3.89 GiB	fits	73.7–147.4 estimated tokens/s
gemma-4-26B-A4B-itQ8_0	Unknown	Needs verification	Estimated speed unavailable
Qwen3-14BQ4_K_M	10.01 GiB	fits	20.4–40.9 estimated tokens/s
gemma-4-31B-itQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
Qwen3-8BQ4_K_M	6.24 GiB	fits	36.6–73.2 estimated tokens/s
Qwen3.8-27BQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
Qwen3.5-4BQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
Qwen3.5-9BQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
Qwen2.5-7B-InstructQ4_K_M	5.58 GiB	fits	39.3–78.6 estimated tokens/s
GLM-5.3-FlashQ8_0	Unknown	Needs verification	Estimated speed unavailable
gpt-oss-20bQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
Qwen2.5-0.5B-InstructQ4_K_M	1.50 GiB	fits	374.4–748.9 estimated tokens/s
DeepSeek-V3.2Q4_K_M	Unknown	Needs verification	Estimated speed unavailable
Llama-3.1-8B-InstructQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
Qwen2.5-VL-7B-InstructQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
Qwen2.5-1.5B-InstructQ4_K_M	2.15 GiB	fits	164.7–329.4 estimated tokens/s
Qwen3-VL-8B-InstructQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
Llama-3.2-1B-InstructQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
Qwen3-1.7BQ4_K_M	2.47 GiB	fits	166.2–332.3 estimated tokens/s
Qwen2.5-3B-InstructQ4_K_M	3.10 GiB	fits	87.4–174.8 estimated tokens/s
Qwen3.6-35B-A3BQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
Qwen3-VL-4B-InstructQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
Qwen3.5-0.8BQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
gpt-oss-120bQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
Qwen3.5-2BQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
Qwen3-32BQ4_K_M	20.40 GiB	fits	9.3–18.6 estimated tokens/s
Qwen3.8-Flash-NextQ8_0	Unknown	Needs verification	Estimated speed unavailable
Qwen3-4B-Instruct-2507Q4_K_M	3.89 GiB	fits	73.7–147.4 estimated tokens/s
Qwen3.6-27BQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
Qwen3-VL-2B-InstructQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
GLM-5.3Q8_0	Unknown	Needs verification	Estimated speed unavailable
Qwen2.5-VL-3B-InstructQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
Qwen3.5-27BQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
Qwen2.5-Coder-14B-InstructQ4_K_M	10.12 GiB	fits	20.5–40.9 estimated tokens/s
Qwen2.5-Coder-7B-InstructQ4_K_M	5.58 GiB	fits	39.3–78.6 estimated tokens/s
SmolVLM2-500M-Video-InstructQ8_0	1.56 GiB	Needs verification	Estimated speed unavailable
Qwen3.5-35B-A3BQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
DeepSeek-R1Q4_K_M	Unknown	Needs verification	Estimated speed unavailable
DeepSeek-R1-Distill-Qwen-1.5BQ4_K_M	2.15 GiB	fits	164.7–329.4 estimated tokens/s
gemma-3-1b-itQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
SmolLM2-135M-InstructQ4_K_M	1.19 GiB	fits	1744.8–3489.7 estimated tokens/s
Qwen2-VL-2B-InstructQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
Llama-3.2-3B-InstructQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
Qwen3-30B-A3B-Instruct-2507Q4_K_M	18.66 GiB	fits	91.7–183.5 estimated tokens/s
Qwen3-30B-A3BQ4_K_M	18.66 GiB	fits	91.7–183.5 estimated tokens/s
gemma-3-4b-itQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
GLM-4.7-FlashQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
GLM-5Q4_K_M	Unknown	Needs verification	Estimated speed unavailable
MiniMax-M2.5Q4_K_M	130.81 GiB	doesn't fit	Estimated speed unavailable
DeepSeek-V3-0324Q4_K_M	Unknown	Needs verification	Estimated speed unavailable
Qwen2.5-14B-InstructQ4_K_M	10.12 GiB	fits	20.5–40.9 estimated tokens/s
GLM-4.5-AirQ4_K_M	69.68 GiB	doesn't fit	Estimated speed unavailable
Qwen2.5-VL-32B-InstructQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
Qwen3.5-122B-A10BQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
SmolLM3-3BQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
Qwen3.5-397B-A17BQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
diffusiongemma-26B-A4B-itQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
granite-4.1-3bQ4_K_M	3.27 GiB	fits	87.6–175.3 estimated tokens/s
Qwen3-4B-Thinking-2507Q4_K_M	3.89 GiB	fits	73.7–147.4 estimated tokens/s
Qwen3-Coder-30B-A3B-InstructQ4_K_M	18.66 GiB	fits	91.7–183.5 estimated tokens/s
GLM-5.1Q8_0	Unknown	Needs verification	Estimated speed unavailable
Qwen2.5-Coder-32B-InstructQ4_K_M	20.49 GiB	fits	9.3–18.5 estimated tokens/s
gemma-2-9b-itQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
Qwen3-Coder-NextQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
Olmo-3-7B-ThinkQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
medgemma-4b-itQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
MiniMax-M2.7Q8_0	228.41 GiB	doesn't fit	Estimated speed unavailable
DeepSeek-Coder-V2-Lite-InstructQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
Qwen2.5-Coder-3B-InstructQ4_K_M	3.10 GiB	fits	87.4–174.8 estimated tokens/s
Qwen3-235B-A22BQ4_K_M	134.13 GiB	doesn't fit	Estimated speed unavailable
Qwen2.5-Coder-1.5B-InstructQ4_K_M	2.15 GiB	fits	164.7–329.4 estimated tokens/s
Meta-Llama-3-8B-InstructQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
DeepSeek-R1-0528-Qwen3-8BQ4_K_M	6.24 GiB	fits	36.6–73.2 estimated tokens/s
phi-4Q4_K_M	10.06 GiB	fits	20.7–41.4 estimated tokens/s
Kimi-K2-InstructQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
gemma-3-27b-itQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
Phi-4-mini-instructQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
DeepSeek-R1-Distill-Qwen-32BQ4_K_M	20.49 GiB	fits	9.3–18.5 estimated tokens/s
DeepSeek-R1-Distill-Qwen-7BQ4_K_M	5.58 GiB	fits	39.3–78.6 estimated tokens/s
gemma-3-12b-itQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
GLM-5.2Q8_0	Unknown	Needs verification	Estimated speed unavailable
granite-3.3-2b-instructQ4_K_M	2.75 GiB	fits	119.1–238.1 estimated tokens/s
Olmo-3-7B-InstructQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
DeepSeek-V3.1Q4_K_M	Unknown	Needs verification	Estimated speed unavailable
DeepSeek-R1-Distill-Qwen-14BQ4_K_M	10.12 GiB	fits	20.5–40.9 estimated tokens/s
Kimi-K2.5Q4_K_M	Unknown	Needs verification	Estimated speed unavailable
Qwen3-VL-30B-A3B-InstructQ4_K_M	18.66 GiB	Needs verification	Estimated speed unavailable
SmolLM2-1.7B-InstructQ4_K_M	2.73 GiB	fits	174.3–348.6 estimated tokens/s
SmolVLM-256M-InstructQ8_0	1.25 GiB	Needs verification	Estimated speed unavailable
Llama-3.3-70B-InstructQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
Qwen3-VL-32B-InstructQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
MiniMax-M3Q8_0	423.93 GiB	Needs verification	Estimated speed unavailable
SmolLM2-360M-InstructQ4_K_M	1.41 GiB	fits	680.0–1360.0 estimated tokens/s
DeepSeek-R1-0528Q4_K_M	Unknown	Needs verification	Estimated speed unavailable
Qwen3-Next-80B-A3B-InstructQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
Meta-Llama-3-8BQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
Llama-4-Scout-17B-16E-InstructQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
Qwen3-235B-A22B-Instruct-2507Q4_K_M	134.13 GiB	doesn't fit	Estimated speed unavailable
granite-vision-4.1-4bQ4_K_M	3.27 GiB	Needs verification	Estimated speed unavailable
MiniMax-M2Q4_K_M	130.81 GiB	doesn't fit	Estimated speed unavailable
DeepSeek-R1-Distill-Llama-8BQ4_K_M	6.08 GiB	fits	37.4–74.8 estimated tokens/s
Llama-3_3-Nemotron-Super-49B-v1_5Q4_K_M	Unknown	Needs verification	Estimated speed unavailable
GLM-4.5Q4_K_M	204.03 GiB	doesn't fit	Estimated speed unavailable
SmolVLM2-2.2B-InstructQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
gemma-3n-E2B-itQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
GLM-4.1V-9B-ThinkingQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
medgemma-1.5-4b-itQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
granite-4.1-8bQ4_K_M	6.61 GiB	fits	34.4–68.8 estimated tokens/s
SmolVLM-500M-InstructQ8_0	1.56 GiB	Needs verification	Estimated speed unavailable
Llama-3.1-70B-InstructQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
GLM-4.6V-FlashQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
granite-3.3-8b-instructQ4_K_M	6.23 GiB	fits	37.2–74.5 estimated tokens/s
granite-4.2-8bQ4_K_M	6.61 GiB	fits	34.4–68.8 estimated tokens/s
GLM-4.7Q4_K_M	204.03 GiB	doesn't fit	Estimated speed unavailable
Qwen2.5-7B-Instruct-1MQ4_K_M	5.58 GiB	fits	39.3–78.6 estimated tokens/s
olmOCR-2-7B-1025Q4_K_M	Unknown	Needs verification	Estimated speed unavailable
gemma-3-270m-itQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
granite-guardian-4.1-8bQ4_K_M	6.39 GiB	fits	36.0–71.9 estimated tokens/s
Qwen3-30B-A3B-Thinking-2507Q4_K_M	18.66 GiB	fits	91.7–183.5 estimated tokens/s
Kimi-K2-ThinkingQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
granite-4.2-3bQ4_K_M	3.40 GiB	fits	82.0–164.0 estimated tokens/s
DeepSeek-R1-Distill-Llama-70BQ4_K_M	41.85 GiB	doesn't fit	Estimated speed unavailable
Phi-4-mini-reasoningQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
granite-4.2-30bQ4_K_M	18.50 GiB	fits	10.4–20.8 estimated tokens/s
Qwen3-Next-80B-A3B-ThinkingQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
SmolVLM2-256M-Video-InstructQ4_K_M	1.21 GiB	Needs verification	Estimated speed unavailable
granite-4.0-h-tinyQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
deepseek-coder-6.7b-instructQ4_K_M	6.80 GiB	fits	45.1–90.1 estimated tokens/s
granite-4.0-h-smallQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
granite-4.0-350mQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
granite-4.0-microQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
GLM-4.6Q4_K_M	203.21 GiB	doesn't fit	Estimated speed unavailable
gpt-oss-safeguard-20bQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
Olmo-3.1-32B-InstructQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
Olmo-3-32B-ThinkQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
DeepSeek-V3.1-TerminusQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
granite-4.0-1bQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
OLMoE-1B-7B-0924-InstructQ4_K_M	5.42 GiB	fits	Estimated speed unavailable
gemma-3n-E4B-itQ4_K_M	Unknown	Needs verification	Estimated speed unavailable
Qwen-AgentWorld-35B-A3BQ8_0	Unknown	Needs verification	Estimated speed unavailable
GLM-4.6VQ4_K_M	67.44 GiB	Needs verification	Estimated speed unavailable

Field-level sources and verification dates

name: Official source · verified 2026-10-22. Manufacturer specification for the named configuration.
memoryGB: Official source · verified 2026-10-22. Manufacturer specification for the named configuration.
bandwidthGBps: Official source · verified 2026-10-22. Manufacturer specification for the named configuration.
powerWatts: Official source · verified 2026-10-22. Sustained AI wall-power draw is not verified. Power-supply ratings and battery charging adapters are not substituted for workload power.
kind: Official source · verified 2026-10-22. Catalogue classification from manufacturer architecture; unified indicates a shared CPU/GPU memory pool.
usableMemoryFraction: Official source · verified 2026-10-22. Editorial planning assumption: reserve 25% of installed memory. Not a manufacturer guarantee or measured GPU allocation. OS, drivers and runtime can impose a lower limit.
hostMemoryGB: Official source · verified 2026-10-22. Unknown separate host-memory configuration. Null also avoids double counting a unified system memory pool.
hostBandwidthGBps: Official source · verified 2026-10-22. Unknown separate host-memory bandwidth. Unified memory is already represented by bandwidthGBps.
gpuCount: Official source · verified 2026-10-22. One physical accelerator or unified SoC represented by this entry.
priceUSD: Official source · verified 2026-10-22. Apple US public Product/Offer JSON-LD read on 2026-10-22. Exact configuration advertised price before tax; a point quote, not a street-price range. Delivery availability was not tested. SKU MHL64LL/A.