Mac Studio M5 Ultra 96 GB
Apple's 2026 desktop workstation. The M5 generation adds Neural Accelerators to every GPU core, which lifts prefill throughput dramatically over M4; for bandwidth-bound decode, achieved memory bandwidth remains one of the strongest predictors once the model fits. Its combination of very large unified memory and low idle power makes it a strong option for running large MoE models at home.
ApplesystemAvailable21 models fit0 measured
Memory
96 GB
Official spec LPDDR5X unified
Memory bandwidth
1,200 GB/s
Official spec ~88% achieved in practice
Specifications
CPU
Apple M5 Ultra, 30-core Official spec
GPU
64-core GPU with Neural Accelerators Official spec
Memory
96 GB LPDDR5X unified Official spec
Memory bandwidth
1,200 GB/s Official spec source
Achieved bandwidth ⓘ
~88% (1,052 GB/s) Assumption
FP16 compute ⓘ
~192 TFLOPS Estimated
Architecture
Unified memory (CPU and GPU share one pool) Official spec
Nominal power
280 W Official spec
Measured load power
190 W Measured
Idle power
12 W Measured
Released
22 Sept 2026 Official spec
Base Ultra configuration (binned 30-core CPU / 64-core GPU, 1 TB SSD) at the published EUR 6,599 list price. 1.2 TB/s is double the M5 Max, which is the entire reason to buy one. Achieved-bandwidth coefficient INHERITED from mac-studio-m5-max-128gb, which shares the same silicon and software stack.
Standout models on this machine
Fastest
gpt-oss-20b — ~261–376 t/s
MXFP4 · estimated confidence
Largest that fits
Compare against
Related hardware
Model performance on Mac Studio M5 Ultra 96 GB0 measured, 21 estimated
21 of 21 rows
| Model↕ | Quant | Memory↕ | Decode▼ | Prefill↕ | Context | Fit | Confidence↕ |
|---|---|---|---|---|---|---|---|
| gpt-oss-20b OpenAI · 20.915B (3.6B active) | MXFP4recommended | 13.0 GB | ~261–376 t/s | ~2680–5560 t/s | 128K | Comfortable | Estimated |
| gpt-oss-120b OpenAI · 116.829B (5.1B active) | MXFP4recommended | 66.5 GB | ~203–293 t/s | ~951–1980 t/s | 128K | Fits | Estimated |
| Nemotron 3.5 Lightning 30B-A3B NVIDIA · 31.578B (3B active) | Q8_0recommended | 33.9 GB | ~180–258 t/s | ~2390–4950 t/s | 1024K | Comfortable | Estimated |
| Qwen3-Coder 30B-A3B Alibaba Qwen · 30.532B (3.3B active) | Q8_0recommended | 33.5 GB | ~167–240 t/s | ~2310–4800 t/s | 256K | Comfortable | Estimated |
| Qwen3 30B-A3B Alibaba Qwen · 30.532B (3.3B active) | Q8_0recommended | 33.5 GB | ~167–240 t/s | ~2310–4800 t/s | 40K | Comfortable | Estimated |
| Mistral Small 4 119B-A6.5B Mistral AI · 119.401B (6.5B active) | Q4_K_Mrecommended | 72.7 GB | ~152–218 t/s | ~833–1730 t/s | 256K | Fits | Estimated |
| Gemma 4 26B-A4B Google DeepMind · 25.806B (3.8B active) | Q8_0recommended | 29.7 GB | ~149–214 t/s | ~2340–4870 t/s | 192K | Comfortable | Estimated |
| Qwen3.5 122B-A10B Alibaba Qwen · 125.086B (10B active) | MLX 4-bitrecommended | 71.4 GB | ~112–162 t/s | ~656–1360 t/s | 256K | Fits | Estimated |
| Nemotron 3 Super 120B-A12B NVIDIA · 123.611B (12B active) | MLX 4-bitrecommended | 70.6 GB | ~95.9–138 t/s | ~603–1250 t/s | 1024K | Fits | Estimated |
| Qwen3 8B Alibaba Qwen · 8.191B | Q8_0recommended | 10.6 GB | ~88.4–127 t/s | ~2830–5890 t/s | 40K | Comfortable | Estimated |
| Phi-4 14B Microsoft · 14.66B | Q8_0recommended | 17.8 GB | ~51.9–74.6 t/s | ~1580–3290 t/s | 16K | Comfortable | Estimated |
| Devstral Small 24B Mistral AI · 23.572B | Q8_0recommended | 26.8 GB | ~33–47.5 t/s | ~985–2050 t/s | 128K | Comfortable | Estimated |
| Mistral Small 3.2 24B Mistral AI · 24.011B | Q8_0recommended | 27.2 GB | ~32.5–46.7 t/s | ~967–2010 t/s | 128K | Comfortable | Estimated |
| Gemma 3 27B Google DeepMind · 27.432B | Q8_0recommended | 33.4 GB | ~28.6–41.1 t/s | ~846–1760 t/s | 96K | Comfortable | Estimated |
| Qwen3.8 27B Alibaba Qwen · 27.781B | Q8_0recommended | 30.4 GB | ~28.2–40.6 t/s | ~836–1740 t/s | 256K | Comfortable | Estimated |
| Qwen3.6 27B Alibaba Qwen · 27.781B | Q8_0recommended | 30.4 GB | ~28.2–40.6 t/s | ~836–1740 t/s | 256K | Comfortable | Estimated |
| Granite 4.2 30B IBM · 29.277B | Q8_0recommended | 33.4 GB | ~26.8–38.6 t/s | ~793–1650 t/s | 128K | Comfortable | Estimated |
| Gemma 4 31B Google DeepMind · 31.273B | Q8_0recommended | 41.0 GB | ~25.2–36.2 t/s | ~742–1540 t/s | 32K | Comfortable | Estimated |
| Qwen3 32B Alibaba Qwen · 32.762B | Q8_0recommended | 37.1 GB | ~24–34.6 t/s | ~709–1470 t/s | 40K | Comfortable | Estimated |
| DeepSeek-R1-Distill 32B DeepSeek · 32.764B | Q8_0recommended | 37.1 GB | ~24–34.6 t/s | ~709–1470 t/s | 128K | Comfortable | Estimated |
| Llama 3.3 70B Meta · 70.554B | Q5_K_Mrecommended | 53.1 GB | ~16.8–24.2 t/s | ~329–683 t/s | 64K | Comfortable | Estimated |
Rows in grey are estimates from our bandwidth model, not measurements — they are shown as a range and never as a precise figure. Use the “measured only” filter to see just the 0 pairings on this machine that a real benchmark backs.
Submit a benchmarkContributions are reviewed before publication.
Ownership economicsUnited States (federal) · C corporation · 8h/day
Monthly economic cost
$62
Calculated after tax
Codex/Claude Code
$100/mo
≈ $100/mo · local is $38 less
Net cash at purchase
$5,945
Calculated VAT not reclaimable
Total over 5 years
$3,741
Calculated after tax, after resale
Cost per USD/1M tokens
$1.54
Calculated 40.4M tokens/month
The $100 comparison uses the Codex Pro 5x / Claude Max 5x plans. A fixed planning conversion is used for USD. This compares monthly spend only: subscriptions have usage limits, local hardware has different capabilities and constraints, and taxes or regional pricing may change the charged amount. Prices checked 7 September 2026.
Purchase-price input: $5,945 in United States (federal), tax/VAT excluded · estimated · Apple Store NL (converted catalogue estimate). This is the localized purchase input; the after-tax economic cost is calculated separately below.
Monthly breakdown
| Depreciation 4,161 over 5 years, straight-line to a 1,783 residual | $69.35 |
| Electricity 8.4 kWh/month at 0.140/kWh | $1.17 |
| Cost of capital 4.0%/yr on 3,864 average capital employed | $12.88 |
| Monthly cost before tax | $83.41 |
| Electricity tax shield Running costs are deductible business expenses | −$0.25 |
| First-year expensing §179 (100.0%) | −$20.81 |
| Monthly economic cost after tax | $62.35 |
Three different numbers, deliberately
Cash cost
$5,945
Money that leaves the bank account on day one, net of reclaimable VAT.
Accounting depreciation
$69.35/month
$4,161 written down over 5 years to a $1,783 residual.
After-tax economic cost
$62.35/month
Depreciation plus running costs plus cost of capital, less the tax those deductions save. This is the figure to compare between machines.
Assumptions and sources (verified 2026-09-07)
VAT rate
0% Official spec
VAT recoverable
0% Official spec
Federal C-corporation estimate at the flat 21% rate. Pass-through entities should use the sole-proprietor estimate as a rougher proxy.
Effective deduction rate
21.00% Calculated
Headline marginal rate 21.00%.
Depreciation
5 years, straight-line Assumption
Residual value
$1,783 (30%) Assumption
Extrapolated from the three-year curve.
Electricity
$0.140/kWh Assumption
Average power draw
48 W Calculated
Load 190 W for 20% of powered hours, idle 12 W for the rest — a machine that is on is not generating tokens the whole time.
Investment allowances
§179 100.0% Official spec
Worth $1,248 in total — first-year expensing that replaces later tax depreciation.
Cost of capital
$12.88/month Assumption
- VAT rate: No US federal VAT · verified 2026-09-07
- Marginal tax rate: IRS Publication 542 — Corporations · verified 2026-09-07
- VAT recoverable fraction: site assumption · verified 2026-09-07
- Useful life: IRS Publication 946 — How To Depreciate Property · verified 2026-09-07
- Electricity price: Site assumption · verified 2026-09-07
Assumes the machine generates tokens 20% of its 176 powered hours per month. Cost per token scales inversely with this number — halve the utilisation and the cost per token doubles.
Federal planning estimate only, not tax advice. State and local income tax, sales/use tax and incentives are excluded. It assumes 100% business use, a Section 179 election, enough business income to use it, and that the full annual limit remains available.