Medium
Medium AI hosting: one GPU for open models and training
One GPU: run 7–70B models, or train your own model for one field.
Last updated
- Server
- One GPU with 24–141 GB of memory
- From
- €0.58 or $1.57 an hour (about €423 or $1,146 a month)
- Checked
- 25 Oct 2026
Monthly costs for hourly prices assume the server runs all month (730 hours).
Medium plans give you one graphics card (GPU) of your own, rented by the hour or by the month. A card with 24 to 32 GB of memory fits models of 7 to 14 billion parameters; models of 27 to 70 billion parameters need 48 to 141 GB.
You can use the same cards to fine-tune a model for your own field. If answering from your own documents is enough, a document chatbot needs no training at all.
Bigger models need several GPUs in one machine: see Large AI hosting. An agent that only calls an online model needs no GPU: see Small AI hosting.
What fits
These are the set-ups our finder suggests for this size, each with the reason.
-
One 24–32 GB GPU
7–14B models fit one RTX 4090, L4 or RTX 5090, and so do QLoRA fine-tunes of that size.
-
One 48–141 GB GPU
27–70B models need 48–141 GB of GPU memory. One 80 GB H100 runs Llama 3.3 70B at 4 bits, and fits a QLoRA fine-tune of 70B.
-
One 48–80 GB GPU, or no training
Most field-specific fine-tunes are QLoRA runs on one 48–80 GB GPU. If answering from your own documents is enough, a document chatbot needs no training at all.
Medium plans we compare
You can filter them by billing or data-centre location. Each price shows the date we last checked it on the provider's own page.
| Provider | What you rent | Size | From | Billing | New-customer offer | Link |
|---|---|---|---|---|---|---|
| Infomaniak | AI Services: mistralai/Mistral-Small-4-119B-2603 Pay-as-you-go API access to Mistral Small 4 (119B) through an OpenAI-compatible interface, run in Infomaniak's own Swiss data centres. | Medium | €0.20 / €0.75 per 1M tokens, input / output · pay as you go, no commitment; billed at the end of each month checked 25 Oct 2026 | Per use | No offer | Visit |
| UpCloud | GPU Server, 1 x NVIDIA L4, 8 cores, 64 GB · 24 GB GPU One dedicated NVIDIA L4 with 8 cores and 64 GB of RAM, an AI-ready Ubuntu template with NVIDIA drivers, charged by the hour in Helsinki. Fits 7–14B models | Medium | €0.58/hour (about €423/month) on demand, billed by the starting hour; monthly cost capped at 28 days of use checked 25 Oct 2026 | Hourly | No offer | Visit |
| UpCloud | GPU Server, 1 x NVIDIA L40S, 8 cores, 64 GB · 48 GB GPU One dedicated NVIDIA L40S with 8 cores and 64 GB of RAM, an AI-ready Ubuntu template with NVIDIA drivers, charged by the hour in Helsinki. | Medium | €1.11/hour (about €810/month) on demand, billed by the starting hour; monthly cost capped at 28 days of use checked 25 Oct 2026 | Hourly | No offer | Visit |
| UpCloud | GPU Server, 1 x NVIDIA H100, 12 cores, 240 GB · 80 GB GPU One dedicated NVIDIA H100 SXM with 12 cores and 240 GB of RAM, an AI-ready Ubuntu template with NVIDIA drivers, charged by the hour in Helsinki. Fits 70B models at 4 bits | Medium | €2.17/hour (about €1,584/month) on demand, billed by the starting hour; monthly cost capped at 28 days of use checked 25 Oct 2026 | Hourly | No offer | Visit |
| DigitalOcean | GPU Droplet, 1 x NVIDIA L40S · 48 GB GPU One NVIDIA L40S with 48 GB of GPU memory, 8 vCPUs and 64 GiB of RAM, priced by the hour and billed by the second. | Medium | $1.57/hour (about $1,146/month) on demand, billed per second · VAT not stated checked 25 Oct 2026 | Per second | New accounts get a $5 credit that lasts 90 days and covers every product except SaaS add-ons checked 26 Oct 2026 | Visit |
| Liquid Web | GPU server, NVIDIA H100 NVL 94GB · 94 GB GPU Dedicated single-tenant server with one NVIDIA H100 NVL (94 GB) and 256 GB of RAM, AI frameworks pre-installed, billed monthly. | Medium | $2,174.25/month monthly; 25% off the first month, then $2,899 · VAT not stated checked 25 Oct 2026 | Monthly | No offer | Visit |
| Liquid Web | GPU server, NVIDIA L40S 48GB · 48 GB GPU Dedicated single-tenant server with one NVIDIA L40S and 256 GB of RAM, AI frameworks such as PyTorch pre-installed, billed monthly. | Medium | $1,049.25/month monthly; 25% off the first month, then $1,399 · VAT not stated checked 25 Oct 2026 | Monthly | No offer | Visit |
| Unihost | AZM-10, NVIDIA H200 · 141 GB GPU A dedicated server in the Netherlands with one H200 (141 GB), an EPYC 9554P processor, 512 GB of DDR5 memory, 8 TB of SSD storage, a 10 Gbps link and 10 TB of monthly traffic. | Medium | €2,999.90/month monthly, paid in advance; minimum one month checked 25 Oct 2026 | Monthly | No offer | Visit |
| Liquid Web | GPU server, NVIDIA L4 24GB · 24 GB GPU Dedicated single-tenant server with one NVIDIA L4 and 128 GB of RAM, AI frameworks such as PyTorch pre-installed, billed monthly. Fits 7–14B models | Medium | $584.25/month monthly; 25% off the first month, then $779 · VAT not stated checked 25 Oct 2026 | Monthly | No offer | Visit |
| GPU Mart | Enterprise Dedicated GPU Server - RTX 5090 · 32 GB GPU A whole server with one RTX 5090 (32 GB), two 18-core E5-2697v4 processors, 256 GB of memory, and SSD, NVMe and SATA drives. Fits 7–14B models | Medium | $479/month monthly rate of a 24-month term, paid in advance · VAT not stated checked 25 Oct 2026 | Monthly | A free 24-hour trial of a VPS, dedicated or GPU server for new clients, on request checked 25 Oct 2026 | Visit |
| GPU Mart | Enterprise Dedicated GPU Server - RTX 4090 · 24 GB GPU A whole server with one RTX 4090 (24 GB), two 18-core E5-2697v4 processors, 256 GB of memory, and SSD, NVMe and SATA drives. Fits 7–14B models | Medium | $409/month monthly rate of a 24-month term, paid in advance · VAT not stated checked 25 Oct 2026 | Monthly | A free 24-hour trial of a VPS, dedicated or GPU server for new clients, on request checked 25 Oct 2026 | Visit |
| GPU Mart | Enterprise Dedicated GPU Server - A100(80GB) · 80 GB GPU A whole server with one 80 GB A100, two 18-core E5-2697v4 processors, 256 GB of memory, and SSD, NVMe and SATA drives. Fits 70B models at 4 bits | Medium | $1,559/month monthly rate of a 24-month term, paid in advance · VAT not stated checked 25 Oct 2026 | Monthly | A free 24-hour trial of a VPS, dedicated or GPU server for new clients, on request checked 25 Oct 2026 | Visit |
| Infomaniak | AI Services: swiss-ai/Apertus-v1.5-70B Pay-as-you-go API access to the Swiss open model Apertus 70B through an OpenAI-compatible interface, run in Infomaniak's own Swiss data centres. | Medium | €0.70 / €2.50 per 1M tokens, input / output · pay as you go, no commitment; billed at the end of each month checked 25 Oct 2026 | Per use | 1M free credits to start with AI Services, then pay as you go checked 25 Oct 2026 | Visit |
| DigitalOcean | GPU Droplet, 1 x NVIDIA H100 · 80 GB GPU One NVIDIA H100 with 80 GB of GPU memory, 20 vCPUs and 240 GiB of RAM, priced by the hour and billed by the second. Fits 70B models at 4 bits | Medium | $4.41/hour (about $3,219/month) on demand, billed per second · VAT not stated checked 25 Oct 2026 | Per second | New accounts get a $5 credit that lasts 90 days and covers every product except SaaS add-ons checked 26 Oct 2026 | Visit |
| GPU Mart | Enterprise Dedicated GPU Server - H100 · 80 GB GPU A whole server with one 80 GB H100, two 18-core E5-2697v4 processors, 256 GB of memory, and SSD, NVMe and SATA drives. Fits 70B models at 4 bits | Medium | $2,099/month monthly rate of a 24-month term, paid in advance · VAT not stated checked 25 Oct 2026 | Monthly | A free 24-hour trial of a VPS, dedicated or GPU server for new clients, on request checked 25 Oct 2026 | Visit |
| Unihost | AND-G5, Nvidia RTX 4090 · 1 GPU A dedicated server in Poland with one RTX 4090, a Xeon Gold 6314U processor, 128 GB of memory, two 960 GB SSDs and a 1 Gbps port with 30 TB of traffic. | Medium | €459.90/month monthly, paid in advance; minimum one month checked 25 Oct 2026 | Monthly | No offer | Visit |
| Unihost | AZM-04, NVIDIA RTX 4090 · 1 GPU A dedicated server in the Netherlands with one RTX 4090, a Xeon E5-2609v4 processor, 64 GB of memory, a 480 GB SSD and a 1 Gbps port with 10 TB of traffic. | Medium | €359.90/month monthly, paid in advance; minimum one month checked 25 Oct 2026 | Monthly | No offer | Visit |
| Unihost | LWG-G13, H100 · 1 GPU A dedicated server in Frankfurt with one H100, two EPYC 7413 processors, 192 GB of memory, two 1.92 TB NVMe drives and a 1 Gbps port with 30 TB of traffic. | Medium | €1,658.90/month monthly, paid in advance; minimum one month; plus a one-off set-up fee of €36.90 checked 25 Oct 2026 | Monthly | No offer | Visit |
| No plan matches these filters. Show all plans | ||||||
Not shown: 0 plans whose data-centre locations we have not confirmed yet.
Where to go next
Not sure which size you need? Answer four questions in our finder.