Does Ternary Bonsai 2 27B fit on a RTX 5090 workstation (1x 32 GB)?

Yes. At PTQ1_0 it needs about 7 GB including the KV cache and runtime overhead, against roughly 29 GB usable — a comfortable fit.

How fast is Ternary Bonsai 2 27B on a RTX 5090 workstation (1x 32 GB)?

121 t/s of decode throughput, from 1 published measurement (low confidence).

What does it cost to run Ternary Bonsai 2 27B on a RTX 5090 workstation (1x 32 GB)?

The default US federal estimate is about $46 per month as a C corporation at 8 hours a day, including depreciation, electricity and tax relief, or roughly $2.98 per million output tokens at 20% generation utilisation. Use the selector for an interactive country estimate.

Ternary Bonsai 2 27B on RTX 5090 workstation (1x 32 GB)

27.36B on 32 GB at 1,792 GB/s. Hardware details · Model details

Yes. Ternary Bonsai 2 27B fits on RTX 5090 workstation (1x 32 GB) at the recommended PTQ1_0 configuration, requiring approximately 7.2 GB at 8K context. Practical context capacity is 256K. Measured evidence: 120.50 t/s median across 1 comparable run at PTQ1_0 and 0 context. Measured

CompatibilityComfortable

Calculated fit

Yes

Calculated Uses under 80% of usable memory. Room for a long context and other work.

Recommended quantization

PTQ1_0

Calculated gguf

Memory required

7.2 GB

Calculated of 29.4 GB usable — 24%

Max practical context

256K

Calculated model supports 256K

Observed resident fit: PTQ1_0 with eng_prism_llama_cpp at 0 context; PQ2_0 with eng_prism_llama_cpp at 0 context. This is measured evidence for those exact artifacts and settings; it does not change the standard memory calculation above.

Memory budget at 8K context

Model weights ⓘ

5.5 GB Measured

KV cache ⓘ

0.5 GB Calculated

Runtime overhead ⓘ

1.2 GB Estimated

Total required

7.2 GB Calculated

Headroom ⓘ

22.2 GB Calculated

Smallest published ternary pack that leaves headroom: uses 24% of usable memory at 8K context.

Discrete GPU: 32 GB of VRAM, of which we assume 92% is usable after driver and context overhead.

Hybrid attention: token-growing KV memory is charged to 16 full-attention blocks. The linear-attention blocks also hold fixed-size state, so the memory estimate is approximate.

Performance

Measured performance · recommended `PTQ1_0`

Decode, prefill and TTFT below are measured for PTQ1_0, the recommended configuration.

Low1/10

Measured decode

121 t/s

Measured single measurement

Measured prefill

1810 t/s

Measured single measurement

TTFT at 8K

~1.8–3.8 s

Estimated PTQ1_0; calculated range

Hardware load reference

640 W

Measured machine-level load; not this model run

Measurement evidence

2 measured runs

Rows are kept separate when quantization, runtime, context, workload or execution settings differ.

Quant	Engine	Decode mode	Context	Runs	Median decode	Median prefill	Latest	Sources
`PQ2_0`	Prism ML llama.cpp fork	standard	0	1	129.90 t/s	3,893 t/s	—	huggingface.co
`PTQ1_0`	Prism ML llama.cpp fork	standard	0	1	120.50 t/s	1,805 t/s	—	huggingface.co

How the estimate is calculated

Decode: reading 27.4B active parameters at 1.75 bits/weight takes 4.22 ms at 1792 GB/s x 79% achieved efficiency.
Prefill: 210 TFLOPS (FP16) x 4 for native FP4 tensor cores x 0.71 calibrated against measured prefill on this platform x 32% assumed model-FLOPs utilisation, divided by 2 x 27.4B 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.

Why this is rated "Low" confidence

Vendor-published benchmark: +1
1 measurement: +0
Single measurement — agreement cannot be assessed: +0
Incomplete run metadata: +0

Score 1/10. Aggregated with the median, not the mean, so one outlier cannot move the headline figure.

This model at other quantizations on this machine

Quantization	Resident	Disk	Total needed	Utilisation	Fit	Max context
`PTQ1_0`recommended	5.5 GB	5.5 GB	7.2 GB	24%	ComfortableResident run observed	256K
`PQ2_0`	6.7 GB	6.7 GB	8.4 GB	29%	ComfortableResident run observed	256K

Cost of running Ternary Bonsai 2 27B on this machineUnited States (federal) · C corporation · 8h/day

Monthly economic cost

$46

Calculated after tax

Codex/Claude Code

$100/mo

≈ $100/mo · local is $54 less

Net cash at purchase

$3,332

Calculated VAT not reclaimable

Total over 5 years

$2,731

Calculated after tax, after resale

Cost per USD/1M tokens

$2.98

Calculated 15.3M 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: $3,332 in United States (federal), tax/VAT excluded · estimated · NL retail (system build) (converted catalogue estimate). This is the localized purchase input; the after-tax economic cost is calculated separately below.

Monthly breakdown

Depreciation 2,832 over 5 years, straight-line to a 500 residual	$47.21
Electricity 32.4 kWh/month at 0.140/kWh	$4.53
Cost of capital 4.0%/yr on 1,916 average capital employed	$6.39
Monthly cost before tax	$58.13
Electricity tax shield Running costs are deductible business expenses	−$0.95
First-year expensing §179 (100.0%)	−$11.66
Monthly economic cost after tax	$45.51

Three different numbers, deliberately

Cash cost

$3,332

Money that leaves the bank account on day one, net of reclaimable VAT.

Accounting depreciation

$47.21/month

$2,832 written down over 5 years to a $500 residual.

After-tax economic cost

$45.51/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

$500 (15%) Assumption

Two GPU generations later, the card is worth a fraction of its list price.

Electricity

$0.140/kWh Assumption

Average power draw

184 W Calculated

Load 640 W for 20% of powered hours, idle 70 W for the rest — a machine that is on is not generating tokens the whole time.

Investment allowances

§179 100.0% Official spec

Worth $700 in total — first-year expensing that replaces later tax depreciation.

Cost of capital

$6.39/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

36.2 t/s

Calculated purchase price only

Decode per $100/month

264.8 t/s

Calculated after-tax ownership cost

Tokens per joule

0.65

Calculated same as tokens/s per watt

USD per 1M output tokens

$2.98

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	29.6M/mo	$24	API cheaper	Different model
DeepSeek-V4.1 Flash DeepSeek	$0.60	25.3M/mo	$27	API cheaper	Different model
DeepSeek-V4 Pro DeepSeek	$1.98	6.3M/mo	$111	Local cheaper	Different model
DeepSeek V4 Pro (Together) Together AI	$4.40	2.1M/mo	$324	Local cheaper	Different model
GLM-5.3 Z.ai	$4.40	2.9M/mo	$238	Local cheaper	Different model
Claude Haiku 4.5 Anthropic	$5.00	3.5M/mo	$199	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 $46/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.

Ternary Bonsai 2 27B on other hardware

Hardware	Decode	Price	Fit
Used RTX 4090 workstation (24 GB)	91.1 t/s	$2,071 *	Comfortable

Nearest alternatives to the RTX 5090 workstation (1x 32 GB)

Ternary Bonsai 2 27B on Dual RTX 5090 workstation (2x 32 GB)64 GB Ternary Bonsai 2 27B on Quad RTX 5090 workstation (4x 32 GB)128 GB Ternary Bonsai 2 27B on Dual used RTX 4090 workstation (48 GB)48 GB Ternary Bonsai 2 27B on Used RTX 4090 workstation (24 GB)24 GB Ternary Bonsai 2 27B on RTX 3090 workstation (Ryzen 9 7950X, 64 GB)24 GB

Compare RTX 5090 workstation (1x 32 GB) against Dual RTX 5090 workstation (2x 32 GB) →