Llama 3.3 70B on RTX PRO 6000 Max-Q workstation (96 GB, 300 W)
70.554B on 96 GB at 1,792 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
Q4_K_M
Calculated gguf
Memory required
45.8 GB
Calculated of 88.3 GB usable — 52%
Max practical context
96K
Calculated model supports 128K
Memory budget at 8K context
Model weights ⓘ
41.0 GB Calculated
KV cache ⓘ
2.5 GB Calculated
Runtime overhead ⓘ
2.2 GB Estimated
Total required
45.8 GB Calculated
Usable memory ⓘ
88.3 GB Assumption
Headroom
42.6 GB Calculated
Highest-precision quantization that leaves headroom: uses 52% of usable memory at 8K context.
Discrete GPU: 96 GB of VRAM, of which we assume 92% is usable after driver and context overhead.
PerformanceEstimated0/10
Decode (generation)
~24.9–35.8 t/s
Estimated calculated, not measured
Prefill (prompt)
~378–784 t/s
Estimated
TTFT at 8K
~10.5–21.8 s
Estimated time to first token
Power while generating
390 W
Measured 0.27 tokens/s per watt
These figures are estimated, not measured. Nobody has published a benchmark of Llama 3.3 70B 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 70.6B active parameters at 4.85 bits/weight takes 32.12 ms at 1792 GB/s x 74% achieved efficiency.
- Prefill: 190 TFLOPS (FP16) x 2 for native FP8 tensor cores x 0.67 calibrated against measured prefill on this platform x 32% assumed model-FLOPs utilisation, divided by 2 x 70.6B parameters per token.
- 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 | 38.4 GB | 43.1 GB | 49% | Comfortable | 96K |
Q4_K_Mrecommended | 41.0 GB | 45.8 GB | 52% | Comfortable | 96K |
Q8_0 | 71.2 GB | 76.8 GB | 87% | Fits | 16K |
BF16 | 131.4 GB | 138.9 GB | 157% | Does not fit | 0 |
Cost of running Llama 3.3 70B on this machineUnited States (federal) · C corporation · 8h/day
Monthly economic cost
$138
Calculated after tax
Codex/Claude Code
$100/mo
≈ $100/mo · local is $38 more
Net cash at purchase
$10,809
Calculated VAT not reclaimable
Total over 5 years
$8,294
Calculated after tax, after resale
Cost per USD/1M tokens
$35.91
Calculated 3.8M 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 9,188 over 5 years, straight-line to a 1,621 residual | $153.13 |
| Electricity 20.1 kWh/month at 0.140/kWh | $2.81 |
| Cost of capital 4.0%/yr on 6,215 average capital employed | $20.72 |
| Monthly cost before tax | $176.65 |
| Electricity tax shield Running costs are deductible business expenses | −$0.59 |
| First-year expensing §179 (100.0%) | −$37.83 |
| Monthly economic cost after tax | $138.23 |
Three different numbers, deliberately
Cash cost
$10,809
Money that leaves the bank account on day one, net of reclaimable VAT.
Accounting depreciation
$153.13/month
$9,188 written down over 5 years to a $1,621 residual.
After-tax economic cost
$138.23/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
$1,621 (15%) Assumption
Two GPU generations later, the card is worth a fraction of its list price.
Electricity
$0.140/kWh Assumption
Average power draw
114 W Calculated
Load 390 W for 20% of powered hours, idle 45 W for the rest — a machine that is on is not generating tokens the whole time.
Investment allowances
§179 100.0% Official spec
Worth $2,270 in total — first-year expensing that replaces later tax depreciation.
Cost of capital
$20.72/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
2.8 t/s
Calculated purchase price only
Decode per $100/month
22.0 t/s
Calculated after-tax ownership cost
Tokens per joule
0.27
Calculated same as tokens/s per watt
USD per 1M output tokens
$35.91
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 | 89.8M/mo | $6 | API cheaper | Different model |
| DeepSeek-V4 Flash DeepSeek | $0.66 | 57.1M/mo | $9 | API cheaper | Different model |
| DeepSeek-V4 Pro DeepSeek | $1.98 | 19.0M/mo | $28 | API cheaper | Different model |
| DeepSeek V4 Pro (Together) Together AI | $4.40 | 6.5M/mo | $82 | API cheaper | Different model |
| GLM-5.3 Z.ai | $4.40 | 8.9M/mo | $60 | API cheaper | Different model |
| Claude Haiku 4.5 Anthropic | $5.00 | 10.6M/mo | $50 | API 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 $138/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.
Llama 3.3 70B on other hardware
| Hardware | Decode | Price | Fit |
|---|---|---|---|
| Dual RTX 5090 workstation (2x 32 GB) | ~38.4–55.2 t/s | €6.499 | Comfortable |
| Quad RTX 5090 workstation (4x 32 GB) | ~32.3–46.4 t/s | €13.999 | Comfortable |
| RTX PRO 6000 Blackwell workstation (96 GB) | ~24.9–35.8 t/s | €11.499 | Comfortable |
| Mac Studio M5 Ultra 96 GB | ~19.7–28.3 t/s | €6.599 | Comfortable |
| Mac Studio M5 Ultra 512 GB | ~11.3–16.3 t/s | €12.599 | Comfortable |
| Mac Studio M5 Ultra 256 GB | ~11.3–16.3 t/s | €9.599 | Comfortable |
Nearest alternatives to the RTX PRO 6000 Max-Q workstation (96 GB, 300 W)
Llama 3.3 70B on RTX PRO 6000 Blackwell workstation (96 GB)96 GBLlama 3.3 70B on Quad RTX 5090 workstation (4x 32 GB)128 GBLlama 3.3 70B on Dual RTX 5090 workstation (2x 32 GB)64 GBLlama 3.3 70B on Dual used RTX 4090 workstation (48 GB)48 GBLlama 3.3 70B on RTX 5090 workstation (1x 32 GB)32 GB
Compare RTX PRO 6000 Max-Q workstation (96 GB, 300 W) against RTX PRO 6000 Blackwell workstation (96 GB) →
Compare RTX PRO 6000 Max-Q workstation (96 GB, 300 W) against RTX PRO 6000 Blackwell workstation (96 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)