Gemma 4 26B-A4B on Mac Studio M3 Ultra 512 GB
25.806B (3.8B active) on 512 GB at 819 GB/s. Hardware details · Model details
CompatibilityComfortable
Fits
Yes
Calculated Uses under 80% of usable memory. Room for a long context and other work.
Recommended quantization
Q8_0
Calculated gguf
Memory required
29.7 GB
Calculated of 435.2 GB usable — 7%
Max practical context
256K
Calculated model supports 256K
Memory budget at 8K context
Model weights ⓘ
26.0 GB Calculated
KV cache ⓘ
1.9 GB Calculated
Runtime overhead ⓘ
1.8 GB Estimated
Total required
29.7 GB Calculated
Usable memory ⓘ
435.2 GB Assumption
Headroom
405.5 GB Calculated
Highest-precision quantization that leaves headroom: uses 7% of usable memory at 8K context.
Unified memory: weights, KV cache and the OS share the same 512 GB pool. We assume 85% is addressable by the GPU, which matches a headless machine with the wired-memory limit raised.
Mixture of experts: all 25.806B parameters must be resident in memory even though only ~3.8B are active per token. Memory follows total parameters; speed follows active parameters.
PerformanceEstimated0/10
Decode (generation)
~89.2–128 t/s
Estimated calculated, not measured
Prefill (prompt)
~707–1470 t/s
Estimated
TTFT at 8K
~5.6–11.7 s
Estimated time to first token
Power while generating
205 W
Measured 2.12 tokens/s per watt
These figures are estimated, not measured. Nobody has published a benchmark of Gemma 4 26B-A4B on this machine that we know of, so we calculate throughput from memory bandwidth and active parameter count. Estimates are always shown as a range. Expand the working below to see exactly how the number was produced.
How the estimate is calculated
- Decode: reading 3.8B active parameters at 8.50 bits/weight takes 6.96 ms at 819 GB/s x 71% achieved efficiency.
- MoE routing penalty of 15% applied: expert gathers are less bandwidth-efficient than a dense sweep.
- Prefill: 57 TFLOPS (FP16) x 1.72 calibrated against measured prefill on this platform x 22% assumed model-FLOPs utilisation, divided by 2 x 9.9B parameters per token.
- Prefill uses 9.9B effective parameters, not the 3.8B active in decode: a batch of hundreds of tokens routes across most of the expert pool.
- MoE decode depends on how well the engine batches expert gathers; real results vary more than for dense models.
- Prefill throughput is highly engine-dependent. Flash attention, batch size and quantized KV all move this number substantially, and for sparse mixture-of-experts models it is the least reliable figure we produce.
- This is a calculated estimate, not a measurement. It assumes a single request, a short prompt, no speculative decoding and a warm model already resident in memory.
Estimator version estimator_v1. Stored with every estimated row so old estimates can be regenerated when the model improves.
This model at other quantizations on this machine
| Quantization | Weights | Total needed | Utilisation | Fit | Max context |
|---|---|---|---|---|---|
MLX 4-bit | 14.1 GB | 17.4 GB | 4% | Comfortable | 256K |
Q4_K_M | 15.0 GB | 18.3 GB | 4% | Comfortable | 256K |
Q8_0recommended | 26.0 GB | 29.7 GB | 7% | Comfortable | 256K |
BF16 | 48.1 GB | 52.4 GB | 12% | Comfortable | 256K |
Cost of running Gemma 4 26B-A4B on this machineUnited States (federal) · C corporation · 8h/day
Monthly economic cost
$87
Calculated after tax
Codex/Claude Code
$100/mo
≈ $100/mo · local is $13 less
Net cash at purchase
$8,287
Calculated VAT not reclaimable
Total over 5 years
$5,198
Calculated after tax, after resale
Cost per USD/1M tokens
$6.28
Calculated 13.8M 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.
Monthly breakdown
| Depreciation 5,801 over 5 years, straight-line to a 2,486 residual | $96.68 |
| Electricity 9.0 kWh/month at 0.140/kWh | $1.27 |
| Cost of capital 4.0%/yr on 5,386 average capital employed | $17.95 |
| Monthly cost before tax | $115.90 |
| Electricity tax shield Running costs are deductible business expenses | −$0.27 |
| First-year expensing §179 (100.0%) | −$29.00 |
| Monthly economic cost after tax | $86.63 |
Three different numbers, deliberately
Cash cost
$8,287
Money that leaves the bank account on day one, net of reclaimable VAT.
Accounting depreciation
$96.68/month
$5,801 written down over 5 years to a $2,486 residual.
After-tax economic cost
$86.63/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
$2,486 (30%) Assumption
Extrapolated from the three-year curve.
Electricity
$0.140/kWh Assumption
Average power draw
51 W Calculated
Load 205 W for 20% of powered hours, idle 13 W for the rest — a machine that is on is not generating tokens the whole time.
Investment allowances
§179 100.0% Official spec
Worth $1,740 in total — first-year expensing that replaces later tax depreciation.
Cost of capital
$17.95/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.
Efficiency metrics
Decode per $1,000 spent
13.1 t/s
Calculated purchase price only
Decode per $100/month
125.6 t/s
Calculated after-tax ownership cost
Tokens per joule
2.12
Calculated same as tokens/s per watt
USD per 1M output tokens
$6.28
Calculated at 20% utilisation
Local versus hosted APIs
| Hosted model | USD/1M output | Break-even | API at your volume | Verdict | Comparison type |
|---|---|---|---|---|---|
| Qwen3.8 Flash Alibaba Cloud | $0.42 | 56.3M/mo | $21 | API cheaper | Different model |
| DeepSeek-V4 Flash DeepSeek | $0.66 | 35.8M/mo | $33 | API cheaper | Different model |
| DeepSeek-V4 Pro DeepSeek | $1.98 | 11.9M/mo | $100 | Local cheaper | Different model |
| DeepSeek V4 Pro (Together) Together AI | $4.40 | 4.1M/mo | $292 | Local cheaper | Different model |
| GLM-5.3 Z.ai | $4.40 | 5.6M/mo | $215 | Local cheaper | Different model |
| Claude Haiku 4.5 Anthropic | $5.00 | 6.7M/mo | $179 | Local cheaper | Different model |
Different-model comparison: the hosted model is not the model you would run locally. Treat this as a workload-quality trade-off, not a direct economic equivalence — the frontier model may complete a task in fewer tokens, or complete tasks the local model cannot. Break-even is the monthly output volume at which API spend equals the $87/month economic cost of owning this machine, assuming 8 input tokens per output token. Hosted prices are published in USD. This table is the canonical US default.
Gemma 4 26B-A4B on other hardware
| Hardware | Decode | Price | Fit |
|---|---|---|---|
| RTX 5090 workstation (1x 32 GB) | ~303–435 t/s | €3.699 | Comfortable |
| Quad RTX 5090 workstation (4x 32 GB) | ~281–404 t/s | €13.999 | Comfortable |
| Dual RTX 5090 workstation (2x 32 GB) | ~216–311 t/s | €6.499 | Comfortable |
| Used RTX 4090 workstation (24 GB) | ~209–301 t/s | €2.299 | Comfortable |
| RTX PRO 6000 Blackwell workstation (96 GB) | ~188–270 t/s | €11.499 | Comfortable |
| RTX PRO 6000 Max-Q workstation (96 GB, 300 W) | ~188–270 t/s | €11.999 | Comfortable |
Nearest alternatives to the Mac Studio M3 Ultra 512 GB
Gemma 4 26B-A4B on Mac Studio M3 Ultra 256 GB256 GBGemma 4 26B-A4B on Mac Studio M5 Ultra 512 GB512 GBGemma 4 26B-A4B on Mac Studio M5 Ultra 256 GB256 GBGemma 4 26B-A4B on Mac Studio M5 Ultra 96 GB96 GBGemma 4 26B-A4B on Mac Studio M5 Max 128 GB128 GB
Compare Mac Studio M3 Ultra 512 GB against Mac Studio M3 Ultra 256 GB →
Compare Mac Studio M3 Ultra 512 GB against Mac Studio M3 Ultra 256 GB →
Page quality score (why this page is or is not indexed)
10/13 — indexed. Generated pages are gated so we do not ask a search engine to rank a page with nothing computed to say. The directive is emitted in the page head via the metadata API, not in the body, so it is authoritative.
- Has a memory-fit calculation (+3)
- Has a performance estimate (+2)
- Has a price (+2)
- Has an ownership economics calculation (+2)
- Well connected (18 internal links) (+1)