2× NVIDIA Tesla P40 (24 GB GDDR5 each, 48 GB total) passed through to your VM, with 20 dedicated cores, 80 GB RAM and 1000 GB NVMe: 2× the single-card plan that Strata was measured on, so every card has the RAM and SSD it needs. Measured with Strata across both cards: 47 tok/s, flat to 87K tokens. Driver preinstalled, the cards are yours alone.
| GPU | 2 × NVIDIA Tesla P40 24 GB GDDR5 · 48 GB VRAM total — full PCI passthrough, all 2 cards in your VM |
|---|---|
| CPU | AMD EPYC Gen 2 / Intel Xeon Gold Gen 2 |
| Dedicated Cores | 20 |
| RAM | 80 GB |
| Storage | 1 TB NVMe |
| Platform | KVM |
| Operating Systems | Linux, Windows |
| Bandwidth | Option 1: Unlimited transfer — 250 Mbps port, grows +1 Mbps/day up to 1 Gbps Option 2: 10 Gbps port — 100TB 10Gbps monthly cap (throttled to ~10 Mbps after cap) |
Real numbers measured on our own VMs on two NVIDIA Tesla P40s, not estimated: Strata first (the 125B model this plan is built for), then llama.cpp with smaller models. GPU plans run on Proxmox with full PCI passthrough: the card is yours, not shared and not virtualized.
Specced for Strata · measured on our own VMs, 17–19 Oct 2026
Qwen3.8-Flash-Next · 125B mixture-of-experts frontier model (6B active per word) · ISTA-DASLab GSQ-RCO IQ3_S · 128K context · Strata v0.1.44.1
| Prompt | 3.8K tokens | 35K tokens | 87K tokens |
|---|---|---|---|
| Reading | 443 tok/s | 712 tok/s | 754 tok/s |
| Writing | 47 tok/s | 49 tok/s | 47 tok/s |
How we run Strata on these cards, step by step →
Strata v0.1.44.1 (commit 296b59d2), ISTA-DASLab GSQ-RCO IQ3_S, CUDA 12.8.93, driver 580.178.04, Ubuntu 26.04.1, AMD EPYC 7532 vCPUs. Prompt reading and decode are the engine's own timings, mean of 2 runs per size, 512 tokens out, thinking off.
Two independent AI servers in one VM, each at full speed.
With Strata, one card already runs the 125B model. Extra cards are for more output or two independent servers. Need more speed per card? The V100 is about 2.5× faster on average, with 32 GB.
With Strata, Qwen3.8-Flash-Next (125B) is the headline. With llama.cpp, mixture-of-experts models such as Qwen3.6-35B-A3B, which activate about 3B of their 35B parameters per word, fit one card and write faster than you read. Memory is what the job actually used on one card.
| Model and setup | Card memory used | Context | Speed |
|---|---|---|---|
| Specced for · Strata Qwen3.8-Flash-Next (125B) Strata · ISTA-DASLab GSQ-RCO IQ3_S · 128K context · the whole card plus 40 GB of system RAM holding the model's experts. | whole card + 40 GB RAM | 128K | 39–41 tok/s writing |
| Our pick for documents Qwen3.8-35B-A3B Distill Newest generation · Q4_K_M · a full 128K context (~200 pages) on one card. Thinks briefly, then answers: a median of 5 seconds per answer on our 40-question document test. | 23.1 GB | 128K | up to 60 tok/s writing |
| Fastest writer Qwen3.6-35B-A3B Mixture of experts · UD-Q4_K_S · 64K context. The fastest writer on one card. | 22.2 GB | 64K | up to 73 tok/s writing |
| Fine-tuning Qwen3.5-9B (QLoRA) QLoRA (4-bit) in Unsloth Studio on Windows Server 2022. | 13.8 GB | — | 76 training tok/s |
The Distill reads documents at about 950 tok/s, so a 15-page document is read in about 9 seconds. Both models carry a built-in accelerator (MTP) that lets them write several words per step.
Serving a team? With Qwen3.6-35B-A3B, add -np 4: four people typing at once each get 28 tok/s, and the card's total output rises to 106 tok/s.
Some teams would rather click than type. Our Windows Server 2022 image comes with the P40 driver installed and the current Visual C++ runtime that modern AI tools need. Install the Unsloth Studio desktop app, pick a model and a dataset, and train. We ran it end to end on this image: the loss fell, the result loaded and answered, and the card did the work.
On Windows the P40 is a compute card: CUDA, PyTorch and training, with no display acceleration.
On Ubuntu the driver is already there. Add Docker and the NVIDIA Container Toolkit, download the model once (21.4 GB), then start the server. Nothing to compile.
mkdir -p /opt/models && curl -L -o /opt/models/Qwen3.6-35B-A3B-UD-Q4_K_S.gguf \
https://huggingface.co/unsloth/Qwen3.6-35B-A3B-MTP-GGUF/resolve/main/Qwen3.6-35B-A3B-UD-Q4_K_S.gguf
docker run -d --name llm --gpus all -v /opt/models:/models -p 127.0.0.1:8080:8080 \
ghcr.io/ggml-org/llama.cpp:server-cuda-b10438 \
-m /models/Qwen3.6-35B-A3B-UD-Q4_K_S.gguf \
--host 0.0.0.0 --port 8080 \
-c 65536 -np 1 -t 6 -tb 6 -b 4096 -ub 2048 \
-ngl auto --fit on --fit-target 1024 \
--spec-type draft-mtp --spec-draft-n-max 2 --jinja-ub 2048--spec-draft-n-max 2-m /models/…ggufThe server speaks the OpenAI API on 127.0.0.1:8080/v1: LangChain, LlamaIndex, Open WebUI and Continue work unchanged. Keep it behind your own proxy with authentication.
Figures: llama.cpp on Ubuntu, 1–2 × P40, 6 vCPU / 30 GB per card · Qwen3.8-35B-A3B Distill on llama.cpp b11459, one P40, 128K context, Oct 2026 · Qwen3.8-Flash-Next on 3 × P40, real workload, Oct 2026 · training on Windows Server 2022, Unsloth Studio, 13 Oct 2026.
Two options per VPS, chosen at checkout: (1) Unlimited transfer starting at a 250 Mbps port that grows +1 Mbps/day up to 1 Gbps; or (2) a 10 Gbps port with a per-plan monthly transfer cap (throttled to ~10 Mbps after the cap).
Run Linux or Windows on any VPS — selected at deployment.