Dedicated GPU VPS · Montreal
A KVM server with a whole NVIDIA Tesla card passed through to your VM, and exactly the RAM, SSD and dedicated cores that Strata needs to run Qwen3.8-Flash-Next at full speed on that one card. Not shared, not virtualized, no per-token bill. Every figure below was measured on our own VMs.
Pick your card
Both run Qwen3.8-Flash-Next with Strata on a single card; the V100 writes about twice as fast and reads prompts about three times as fast. These specs are not random. We ran Strata on both cards, found the RAM where Qwen3.8-Flash-Next (125B) runs at full speed (40 GB on the P40, 35 GB on the V100), the SSD it needs (78 GB of model files, a 56 GB prepared copy, swap and the OS) and the cores it actually uses, then built the plans around exactly that. The specs are made to maximize the performance-to-price ratio.
24 GB · full PCI passthrough
Rent a VPS with a whole Tesla P40 passed through to it. Run open models, serve your team, read your documents and fine-tune on your own data. Nothing leaves the VM, and there is no per-token bill.
32 GB HBM2 · full PCI passthrough
Rent a VPS with a whole Tesla V100 passed through to it. 32 GB of fast HBM2 memory and tensor cores run large, high-precision models and read whole documents in seconds. Serve your team, build agents, keep your data private. Nothing leaves the VM, and there is no per-token bill.
All GPU plans
Every plan gets the two bandwidth options and Linux or Windows like any ServaRica VPS. The CPU, RAM and NVMe scale with the card count.
| CPU Type | Unlimited 250Mbps+ | Actions | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P40 GPU Slice 1 - 1xP40 (24GB VRAM) | 1 × NVIDIA Tesla P40 24 GB 10 cores AMD EPYC Gen 2 / Intel Xeon Gold Gen 2 | 40 GB | 500 GB NVMe | — | — | — | — | — | — | — | ✓ | 100TB 10Gbps | $139CA$191/mo $1529/yrCA$2110/yr | Order Details | |
| V100 GPU Slice 1 - 1xV100 (32GB VRAM) | 1 × NVIDIA Tesla V100 32 GB 8 cores AMD EPYC Gen 2 / Intel Xeon Gold Gen 2 | 35 GB | 400 GB NVMe | — | — | — | — | — | — | — | ✓ | 100TB 10Gbps | $179CA$247/mo $1969/yrCA$2717/yr | Order Details | |
| P40 GPU Slice 2 - 2xP40 (48GB VRAM) | 2 × NVIDIA Tesla P40 24 GB 20 cores AMD EPYC Gen 2 / Intel Xeon Gold Gen 2 | 80 GB | 1 TB NVMe | — | — | — | — | — | — | — | ✓ | 100TB 10Gbps | $278CA$383/mo $3058/yrCA$4220/yr | Order Details | |
| V100 GPU Slice 2 - 2xV100 (64GB VRAM) | 2 × NVIDIA Tesla V100 32 GB 16 cores AMD EPYC Gen 2 / Intel Xeon Gold Gen 2 | 70 GB | 800 GB NVMe | — | — | — | — | — | — | — | ✓ | 100TB 10Gbps | $358CA$494/mo $3938/yrCA$5434/yr | Order Details | |
| P40 GPU Slice 3 - 3xP40 (72GB VRAM) | 3 × NVIDIA Tesla P40 24 GB 30 cores AMD EPYC Gen 2 / Intel Xeon Gold Gen 2 | 120 GB | 1.5 TB NVMe | — | — | — | — | — | — | — | ✓ | 100TB 10Gbps | $417CA$575/mo $4587/yrCA$6330/yr | Order Details | |
| V100 GPU Slice 3 - 3xV100 (96GB VRAM) | 3 × NVIDIA Tesla V100 32 GB 24 cores AMD EPYC Gen 2 | 105 GB | 1.2 TB NVMe | — | — | — | — | — | — | — | ✓ | 100TB 10Gbps | $537CA$741/mo $5907/yrCA$8151/yr | Order Details |
No plans match your current filters. Try removing some filters.
Why these cards
We chose these two cards because, against every GPU on the market, they give the best performance for the money, and we shaped each plan around what Strata measured on them. Strata runs Qwen3.8-Flash-Next (125B), a frontier mixture-of-experts model, on a single card. The card holds the attention layers and part of the model's experts; the rest of the experts stay resident in the VM's system RAM and only what each token needs moves to the GPU. That is why system RAM, not CPU cores, decides the speed, and why one card can do what used to take three. GPU plans run on Proxmox with full PCI passthrough: the card is yours, not shared and not virtualized.
It is a 2016 data-center card. We chose it because 24 GB at this price makes a private AI server affordable.
It is a 2017 data-center card, chosen because 32 GB of fast memory and tensor cores make a serious private AI server affordable.