Privacy
Prompts and decisions stay inside your deployment boundary when the machine and service are configured locally.
Pre-installed hardware
A local system, ready for your models.
Compare sourced configurations, see which published model variants fit, and ask us to prepare a quote.
Latency
Inference does not need a network round trip to a hosted model. Measure it on your intended device.
Cost at scale
Compare a known hardware purchase and operating cost with the per-call costs of hosted inference.
Offline and air-gapped
After downloading model files, an isolated system can serve without an external model connection.
Selected model: JPT-4B
NVIDIA GeForce RTX 5090 (32 GB) workstation Compatibility depends on the model variant and the target configuration.
View model requirements →Device classes
Compare sourced hardware
Prices use the median sourced street price in each currency. Indicative quotes add 30%, or at least 150 currency units, and round up to the next amount ending in 9. Shipping and VAT are excluded.
Apple Silicon
Mac mini and Mac Studio configurations. Unified memory is the installed memory for each listed configuration.
Install tested: none on listed devices; no recorded recipe test matches this platform and memory tier. Recipe available but unverified for decider-12b v2, Imajev-4B, Jevify Gemma 4 26B-A4B, NeoHorse Jev 4B, plus 2 more. Listed devices are fit estimates.
Apple Silicon
Apple Mac mini M6 (16 GB, 256 GB)
Mac mini and Mac Studio configurations. Unified memory is the installed memory for each listed configuration.
Memory: 16 GB (unified)
$1,169 indicative
Median sourced street price: $899 before the quote margin.
Price sources (1)
€1,369 indicative
Median sourced street price: €1,049 before the quote margin.
Price sources (1)
Apple Silicon
Apple Mac mini M5 Pro (15-core CPU, 16-core GPU, 24 GB, 512 GB)
Mac mini and Mac Studio configurations. Unified memory is the installed memory for each listed configuration.
Memory: 24 GB (unified)
$2,209 indicative
Median sourced street price: $1,699 before the quote margin.
Price sources (1)
€2,599 indicative
Median sourced street price: €1,999 before the quote margin.
Price sources (1)
Apple Silicon
Apple Mac mini M5 Pro (18-core CPU, 20-core GPU, 64 GB, 1 TB)
Mac mini and Mac Studio configurations. Unified memory is the installed memory for each listed configuration.
Memory: 64 GB (unified)
$2,469 indicative
Median sourced street price: $1,899 before the quote margin.
Price sources (1)
€2,889 indicative
Median sourced street price: €2,219 before the quote margin.
Price sources (1)
Apple Silicon
Apple Mac Studio M5 Max (18-core CPU, 32-core GPU, 36 GB, 512 GB)
Mac mini and Mac Studio configurations. Unified memory is the installed memory for each listed configuration.
Memory: 36 GB (unified)
$3,249 indicative
Median sourced street price: $2,499 before the quote margin.
Price sources (1)
€3,899 indicative
Median sourced street price: €2,999 before the quote margin.
Price sources (1)
Apple Silicon
Apple Mac Studio M5 Max (18-core CPU, 40-core GPU, 64 GB, 1 TB)
Mac mini and Mac Studio configurations. Unified memory is the installed memory for each listed configuration.
Memory: 64 GB (unified)
$4,029 indicative
Median sourced street price: $3,099 before the quote margin.
Price sources (1)
€4,759 indicative
Median sourced street price: €3,659 before the quote margin.
Price sources (1)
Apple Silicon
Apple Mac Studio M5 Ultra (30-core CPU, 64-core GPU, 96 GB, 1 TB)
Mac mini and Mac Studio configurations. Unified memory is the installed memory for each listed configuration.
Memory: 96 GB (unified)
$7,149 indicative
Median sourced street price: $5,499 before the quote margin.
Price sources (1)
€8,579 indicative
Median sourced street price: €6,599 before the quote margin.
Price sources (1)
GPU workstation
Complete desktop systems with a dedicated NVIDIA GPU.
Install tested: recorded checks on H2O-Lightning-4B v1.1 on RunPod RTX4090 24GB and CoreWeave RTX PRO6000 Blackwell96GB, Linux, 2026-10-09. Listed devices are fit estimates; the hardware names above identify the recorded test systems.
GPU card street price plus a base system with CPU, 64–128 GB RAM, 2 TB NVMe, PSU, and case.
GPU workstation
NVIDIA RTX PRO 6000 Blackwell (96 GB) workstation
Complete desktop systems with a dedicated NVIDIA GPU.
Memory: 96 GB (vram)
$25,359 indicative
Sourced GPU price plus an estimated base system: $19,500 before the quote margin.
Price sources (1)
GPU workstation
NVIDIA GeForce RTX 5090 (32 GB) workstation
Complete desktop systems with a dedicated NVIDIA GPU.
Memory: 32 GB (vram)
$12,799 indicative
Sourced GPU price plus an estimated base system: $9,841 before the quote margin.
Price sources (1)
DGX Spark class
A 128 GB NVIDIA GB10 system. Compatibility is checked for each model on request.
Install tested: none on listed devices; no recorded recipe test matches this platform and memory tier. Listed devices are fit estimates.
DGX Spark class
NVIDIA DGX Spark (128 GB)
A 128 GB NVIDIA GB10 system. Compatibility is checked for each model on request.
Memory: 128 GB (unified)
$9,269 indicative
Median sourced street price: $7,125 before the quote margin.
Price sources (2)
DGX Spark class
ASUS Ascent GX10 (GB10, 1 TB SSD)
A 128 GB NVIDIA GB10 system. Compatibility is checked for each model on request.
Memory: 128 GB (unified)
$4,409 indicative
Median sourced street price: $3,384 before the quote margin.
Price sources (1)
DGX Spark class
Dell Pro Max GB10, HP ZGX Nano, Lenovo, MSI GB10 boxes
A 128 GB NVIDIA GB10 system. Compatibility is checked for each model on request.
Memory: 128 GB (unified)
No sourced price is available for this configuration. Ask us to confirm availability and pricing.
Jetson Orin edge
Compact NVIDIA edge systems for small quantised models.
Install tested: none on listed devices; no recorded recipe test matches this platform and memory tier. Listed devices are fit estimates.
Jetson Orin edge
NVIDIA Jetson Orin Nano Super Developer Kit (8 GB)
Compact NVIDIA edge systems for small quantised models.
Memory: 8 GB (unified)
$549 indicative
Median sourced street price: $399 before the quote margin.
Price sources (1)
€499 indicative
Median sourced street price: €348 before the quote margin.
Price sources (1)
Jetson Orin edge
NVIDIA Jetson AGX Orin 64 GB Developer Kit
Compact NVIDIA edge systems for small quantised models.
Memory: 64 GB (unified)
$4,549 indicative
Median sourced street price: $3,499 before the quote margin.
Price sources (1)
Strong Android device
A high-end Android phone for small on-device models.
Install tested: none on listed devices; no recorded recipe test matches this platform and memory tier. Listed devices are fit estimates.
Strong Android device
Samsung Galaxy S26 Ultra (256 GB model)
A high-end Android phone for small on-device models.
Memory: 12 GB (ram)
$1,289 indicative
Median sourced street price: $990 before the quote margin.
Price sources (1)
How we price this
For each currency, we take the median of the sourced street prices, add 30% or at least 150 units, then round upward to an amount ending in 9. For GPU workstations, the street price includes the GPU card plus a USD 1,500 / EUR 1,400 base system: CPU, 64–128 GB RAM, 2 TB NVMe, PSU, and case. The final quote is confirmed by email.
Model fit
Which hardware for which model
Suggestions use an executable recipe and recorded memory requirements. A tested recipe may have been tested on different hardware; each listed device is a fit estimate.
| Model | Smallest fitting option | Indicative price | Install check | Details |
|---|---|---|---|---|
| AutoJev-27B | NVIDIA RTX PRO 6000 Blackwell (96 GB) workstation 96 GB vram | $25,359 | Recipe ready · unverified; this device is a fit estimate | Model requirements → |
| Bobcat Flash 1.2 | NVIDIA RTX PRO 6000 Blackwell (96 GB) workstation 96 GB vram | $25,359 | Recipe ready · unverified; this device is a fit estimate | Model requirements → |
| decider-12b v2 | Apple Mac mini M5 Pro (18-core CPU, 20-core GPU, 64 GB, 1 TB) 64 GB unified | $2,469 | Recipe ready · unverified; this device is a fit estimate | Model requirements → |
| decisio v0.8.0 on gemma-4-31B-it | NVIDIA RTX PRO 6000 Blackwell (96 GB) workstation 96 GB vram | $25,359 | Recipe ready · unverified; this device is a fit estimate | Model requirements → |
| deck-31B | NVIDIA RTX PRO 6000 Blackwell (96 GB) workstation 96 GB vram | $25,359 | Recipe ready · unverified; this device is a fit estimate | Model requirements → |
| H2O-Lightning-4B v1.1 | NVIDIA GeForce RTX 5090 (32 GB) workstation 32 GB vram | $12,799 | Recipe tested on other hardware; this device is a fit estimate | Model requirements → |
| Imajev-4B | Apple Mac mini M6 (16 GB, 256 GB) 16 GB unified | $1,169 | Recipe ready · unverified; this device is a fit estimate | Model requirements → |
| Jev-Omni | NVIDIA RTX PRO 6000 Blackwell (96 GB) workstation 96 GB vram | $25,359 | Recipe ready · unverified; this device is a fit estimate | Model requirements → |
| Jevify Gemma 4 26B-A4B | Apple Mac Studio M5 Ultra (30-core CPU, 64-core GPU, 96 GB, 1 TB) 96 GB unified | $7,149 | Recipe ready · unverified; this device is a fit estimate | Model requirements → |
| JPT-4B | NVIDIA GeForce RTX 5090 (32 GB) workstation 32 GB vram | $12,799 | Recipe ready · unverified; this device is a fit estimate | Model requirements → |
| NeoHorse Jev 4B | Apple Mac mini M5 Pro (15-core CPU, 16-core GPU, 24 GB, 512 GB) 24 GB unified | $2,209 | Recipe ready · unverified; this device is a fit estimate | Model requirements → |
| Quyet-1.0-Large | Apple Mac Studio M5 Ultra (30-core CPU, 64-core GPU, 96 GB, 1 TB) 96 GB unified | $7,149 | Recipe ready · unverified; this device is a fit estimate | Model requirements → |
| reflex 4B | Apple Mac mini M5 Pro (18-core CPU, 20-core GPU, 64 GB, 1 TB) 64 GB unified | $2,469 | Recipe ready · unverified; this device is a fit estimate | Model requirements → |
| René-1 31B FP8 | NVIDIA RTX PRO 6000 Blackwell (96 GB) workstation 96 GB vram | $25,359 | Recipe ready · unverified; this device is a fit estimate | Model requirements → |
| shisa-de-1 | NVIDIA RTX PRO 6000 Blackwell (96 GB) workstation 96 GB vram | $25,359 | Recipe ready · unverified; this device is a fit estimate | Model requirements → |
| Surogate Rune 26B-A4B v3 | NVIDIA RTX PRO 6000 Blackwell (96 GB) workstation 96 GB vram | $25,359 | Recipe ready · unverified; this device is a fit estimate | Model requirements → |
| torchcast-decision-12b | NVIDIA RTX PRO 6000 Blackwell (96 GB) workstation 96 GB vram | $25,359 | Recipe ready · unverified; this device is a fit estimate | Model requirements → |
| Xor 26B-A4B | NVIDIA RTX PRO 6000 Blackwell (96 GB) workstation 96 GB vram | $25,359 | Recipe ready · unverified; this device is a fit estimate | Model requirements → |
Swipe horizontally to see all model details.
Cloud and remote GPUs
Use a GPU you already rent
For each model, this lists the least expensive sourced instance with enough recorded GPU memory. Prices change; check before renting.
| Model | Recommended VRAM | Cheapest matching instance | Price | Source and date |
|---|---|---|---|---|
| AutoJev-27B | 80 GB | RunPod · A100 80GB PCIe secure A100 · 80 GB | $1.79/h | Price source · 2026-10-09 · not stated (API) |
| Bobcat Flash 1.2 | 80 GB | RunPod · A100 80GB PCIe secure A100 · 80 GB | $1.79/h | Price source · 2026-10-09 · not stated (API) |
| decider-12b v2 | 64 GB | RunPod · A100 80GB PCIe secure A100 · 80 GB | $1.79/h | Price source · 2026-10-09 · not stated (API) |
| decisio v0.8.0 on gemma-4-31B-it | 48 GB | RunPod · L40S community L40S · 48 GB | $0.79/h | Price source · 2026-10-09 · not stated (API) |
| deck-31B | 80 GB | RunPod · A100 80GB PCIe secure A100 · 80 GB | $1.79/h | Price source · 2026-10-09 · not stated (API) |
| H2O-Lightning-4B v1.1 | 24 GB | RunPod · RTX 4090 community RTX 4090 · 24 GB | $0.34/h | Price source · 2026-10-09 · not stated (API) |
| Imajev-4B | 12 GB | RunPod · RTX 4090 community RTX 4090 · 24 GB | $0.34/h | Price source · 2026-10-09 · not stated (API) |
| Jev-Omni | 48 GB | RunPod · L40S community L40S · 48 GB | $0.79/h | Price source · 2026-10-09 · not stated (API) |
| Jevify Gemma 4 26B-A4B | 64 GB | RunPod · A100 80GB PCIe secure A100 · 80 GB | $1.79/h | Price source · 2026-10-09 · not stated (API) |
| JPT-4B | 12 GB | RunPod · RTX 4090 community RTX 4090 · 24 GB | $0.34/h | Price source · 2026-10-09 · not stated (API) |
| NeoHorse Jev 4B | 12 GB | RunPod · RTX 4090 community RTX 4090 · 24 GB | $0.34/h | Price source · 2026-10-09 · not stated (API) |
| Quyet-1.0-Large | 80 GB | RunPod · A100 80GB PCIe secure A100 · 80 GB | $1.79/h | Price source · 2026-10-09 · not stated (API) |
| reflex 4B | 12 GB | RunPod · RTX 4090 community RTX 4090 · 24 GB | $0.34/h | Price source · 2026-10-09 · not stated (API) |
| René-1 31B FP8 | 48 GB | RunPod · L40S community L40S · 48 GB | $0.79/h | Price source · 2026-10-09 · not stated (API) |
| shisa-de-1 | 64 GB | RunPod · A100 80GB PCIe secure A100 · 80 GB | $1.79/h | Price source · 2026-10-09 · not stated (API) |
| Surogate Rune 26B-A4B v3 | 80 GB | RunPod · A100 80GB PCIe secure A100 · 80 GB | $1.79/h | Price source · 2026-10-09 · not stated (API) |
| torchcast-decision-12b | 48 GB | RunPod · L40S community L40S · 48 GB | $0.79/h | Price source · 2026-10-09 · not stated (API) |
| Xor 26B-A4B | 64 GB | RunPod · A100 80GB PCIe secure A100 · 80 GB | $1.79/h | Price source · 2026-10-09 · not stated (API) |
Swipe horizontally to see all cloud price and source details.
Availability, region, storage, and egress can change the total cost. Check the provider’s current listing before renting.
Request a setup
Tell us what you need
We will check the requested configuration and reply with an indicative quote. Final pricing depends on availability, shipping, and VAT.