"Run your own ChatGPT for five bucks a month" sounds like cheap
clickbait. And mostly it is. But the gap between clickbait and reality
has narrowed a lot in 2026. Quantized 3B and 7B models have become
genuinely useful. VPS providers now pack 8 GB of RAM into plans under
\$6. If you pick carefully, you can run a respectable local model on a
budget server.
This article is about what actually works, what falls over, and which
provider from our hosting comparison gives you the most
compute per dollar for AI workloads.
The short version
- You will not run GPT-4-class models on a \$5 VPS. Stop expecting to.
- You can run Llama 3.2-3B, Qwen 2.5-7B, or Mistral-7B in 4-bit quantization if you have at least 8 GB of RAM.
- CPU inference is slow (5 to 15 tokens per second) but usable for personal automation, chatbots with low traffic, or embedding generation.
- The providers that matter for this use case: Contabo, Hetzner, and Netcup. Two are based in Germany and the third in Austria, which also helps with EU data residency.
- Avoid US hyperscaler \$5 plans (DO, Vultr, Linode) for LLM work. Same price, one quarter of the RAM.
Why self-host an LLM at all?
Three reasons keep coming up in our research:
Privacy. 44% of organizations cite data privacy as the top barrier
to adopting hosted LLM APIs. Self-hosting means your prompts, documents,
and customer data never leave your box. For law firms, clinics, and
anyone building internal knowledge bases, this is non-negotiable.
Cost at volume. Hosted API costs add up fast once you run batch
jobs, background summarization, or embedding pipelines. The rough
break-even with a \$5 VPS sits around two million tokens per day for
small models. Below that, APIs are cheaper. Above that, self-hosting
starts to win.
Learning and control. You get to pick the model, tune the system
prompt at the weight level (via LoRA adapters), and run without rate
limits. Worth it for developers.
What "self-hosted LLM" actually needs
Before we compare VPS plans, a quick hardware reality check. On CPU-only
servers (which every budget VPS is), the bottleneck is RAM bandwidth,
not CPU speed. Model weights must fit in RAM, and every token generation
pass reads the entire weight file. So the two numbers that matter are:
- RAM size. Determines which model you can load at all.
- RAM bandwidth. Determines how many tokens per second you get.
Here is roughly what each tier can do in 4-bit quantization (GGUF
format, llama.cpp inference):
| RAM | Largest usable model | Realistic tokens/sec (CPU) | Good for |
|---|---|---|---|
| 2 GB | Qwen 2.5-0.5B, Llama 3.2-1B | 20-40 tok/s | Embeddings, simple classification |
| 4 GB | Llama 3.2-3B, Phi-3.5 Mini | 8-15 tok/s | Personal chatbots, summarization |
| 8 GB | Qwen 2.5-7B, Mistral-7B, Llama 3.1-8B | 4-8 tok/s | Production-adjacent use, RAG |
| 16 GB | Mixtral 8x7B (barely), Qwen 2.5-14B | 2-4 tok/s | More complex reasoning |
Below 8 GB you are limited to small models. Above 8 GB you unlock the
"actually useful" tier. This matters enormously when picking a VPS.
The \$3 to \$6 VPS plans that can run real LLMs
Pulled directly from our database, filtered to plans where the RAM is at
least 8 GB (the practical floor for running a 7B model in 4-bit
quantization):
Update (August 2026): prices re-verified against our nightly tracker. Since this article was first published, Hetzner raised the CX33 from €5.49 to €8.49 (+55%) and the CAX21 from €6.49 to €10.49, and Netcup's VPS 1000 ARM G11 went from €6.26 to €7.77. Contabo has not moved. Full change history: Hetzner, Netcup, Contabo.
| Provider | Plan | Price / mo | RAM | vCPU | Storage | Bandwidth |
|---|---|---|---|---|---|---|
| Contabo | Cloud VPS 10 | €3.60 | 8 GB | 4 vCPU | 75 GB NVMe | Unlimited |
| Hetzner | CX33 | €8.49 | 8 GB | 4 vCPU | 80 GB NVMe | 20 TB |
| Contabo | Cloud VPS 20 | €5.60 | 12 GB | 6 vCPU | 100 GB NVMe | Unlimited |
| Netcup | VPS 1000 ARM G11 | €7.77 | 8 GB | 6 vCPU (ARM) | 256 GB NVMe | Unlimited |
| Hetzner | CAX21 (ARM) | €10.49 | 8 GB | 4 vCPU (ARM) | 80 GB NVMe | 20 TB |
Compare that to the "famous" \$5 cloud plans:
| Provider | Plan | Price / mo | RAM | vCPU |
|---|---|---|---|---|
| DigitalOcean | Basic Regular | \$4.00 | 512 MiB | 1 vCPU |
| Vultr | CC Regular | \$2.50 | 512 MB | 1 vCPU |
| Linode | Shared Nanode | \$5.00 | 1 GB | 1 vCPU |
That 8x RAM difference at the same price point is the whole reason
European providers win this category. DO, Vultr and Linode are excellent
for web apps and small services. They are not designed for the RAM-heavy
workload an LLM needs.
Provider deep dive
Contabo: the price-per-GB winner
Contabo Cloud VPS 10 at €3.60 gives you 8 GB RAM, 4 vCPU and 75 GB
NVMe. Nothing else in the market matches this at that price. The catch
is real: Contabo is known for inconsistent CPU steal on noisy neighbors,
and network latency can be variable. For LLM inference (which is
RAM-bound, not network-bound) that tradeoff is acceptable.
Go up one tier to the Cloud VPS 20 at €5.60 for 12 GB RAM and 6 vCPU.
This is the sweet spot for running Qwen 2.5-14B in 4-bit and still
having headroom for a web server and a small vector database on the same
machine.
Hetzner: the quality-per-dollar winner
The Hetzner CX33 at €8.49 with 8 GB RAM and 4 vCPU is the most reliable
choice in this range. AMD EPYC cores, fast NVMe, 20 TB of bandwidth,
datacenters in Germany, Finland and Virginia. Support is competent, the
control panel is minimal in a good way.
The ARM variant (CAX21) at €10.49 is especially interesting for LLM
work. Ampere Altra cores benchmark about 15 to 25 percent faster than
the x86 CX33 on llama.cpp, because of higher sustained memory bandwidth.
Same price range, noticeably better tokens per second.
Netcup: the quiet competitor
Less known outside the German-speaking market. The VPS 1000 ARM G11 at
€7.77 gives you 8 GB RAM, 6 vCPU, and 256 GB of NVMe, which is the
largest disk in this comparison. Useful if you want to store multiple
models and switch between them without re-downloading 4 to 8 GB each
time.
What you will actually run on 8 GB
Here is the honest menu for an 8 GB VPS with Ollama or llama.cpp:
- Qwen 2.5-7B-Instruct (Q4_K_M). Best general-purpose model in this size. Strong reasoning, multi-language, clean code.
- Llama 3.1-8B-Instruct (Q4_K_M). Ubiquitous, well-supported, safe default. Slightly weaker than Qwen on code.
- Mistral-7B-Instruct-v0.3 (Q5_K_M). Older but predictable. Good for summarization.
- Phi-3.5-Mini (Q6_K). Smaller (3.8B), fits in 4 GB of RAM, quick for chat.
- nomic-embed-text or bge-small-en. Embedding models for RAG pipelines. Run these alongside the main model.
What you will not run well: anything 14B+, models with vision
capabilities (qwen-vl, llava), or anything demanding low-latency
streaming to many users.
Practical setup in 15 minutes
Rough sketch so you know what you are signing up for. This assumes
Ubuntu 22.04 or 24.04 on a 8 GB VPS.
# 1. Install Ollama (handles download, quantization, serving)
curl -fsSL https://ollama.com/install.sh | sh
# 2. Pull a model
ollama pull qwen2.5:7b-instruct-q4_K_M
# 3. Test
ollama run qwen2.5:7b-instruct-q4_K_M "Summarize the French Revolution in two paragraphs."
# 4. Expose the API (careful with firewall rules)
# Ollama serves on localhost:11434 by default. Reverse-proxy with Caddy or Nginx.
Add Open WebUI if you want a ChatGPT-like browser interface. Add
llama-swap if you want to load different models on demand without
restarting.
When NOT to self-host
A \$5 VPS gives you about 5 to 8 tokens per second on a 7B model. That
is fine for one user typing in a chat. It is not fine for:
- Real-time voice assistants. Latency will be embarrassing.
- Multi-user chat services. Four concurrent users will queue and feel broken.
- Anything that needs tool use, function calling at scale, or long-context (>8K tokens).
- Production RAG over millions of documents with tight latency SLAs.
For those workloads, you want a GPU instance. That immediately pushes
you above \$100/month with Vast.ai, RunPod, or a dedicated server with a
consumer GPU. At that point, OpenAI API or Anthropic API is often
cheaper up to surprisingly high volumes.
Quick comparison: which \$5 to \$7 VPS wins?
| Use case | Best pick | Why |
|---|---|---|
| Lowest price with 8 GB RAM | Contabo Cloud VPS 10 | €3.60 for 8 GB is unmatched |
| Most reliable inference | Hetzner CX33 | Predictable performance, no noisy neighbors |
| Best tokens per second per dollar | Hetzner CAX21 (ARM) | Higher sustained memory bandwidth |
| Largest model storage | Netcup VPS 1000 ARM G11 | 256 GB NVMe stores a dozen models easily |
| Running LLM + web app on one box | Contabo Cloud VPS 20 | 12 GB RAM leaves room for Postgres and nginx |
You can also compare them side by side on Contabo vs
Hetzner or review the full cloud VPS
category.
What changes this year
Two developments worth tracking in 2026:
Smaller models keep getting smarter. Phi-3.5 Mini already
outperforms last year's 13B models on some benchmarks. Expect 2B and 3B
models to be genuinely production-useful by Q4 2026. That makes the 2 GB
RAM tier (under \$2) a legitimate option.
ARM inference is closing the gap. Apple showed what M-series can do.
Ampere Altra and AWS Graviton are bringing similar memory bandwidth
advantages to \$5 VPS plans. If you have not tried ARM yet, this is the
year.
The verdict
If you want to learn, build a personal assistant, run background
summarization, or stand up a private chatbot for a small team, a \$5 to
\$7 VPS with 8 GB of RAM is genuinely enough in 2026. Pick
Contabo for raw price, Hetzner
for reliability, Netcup for storage. Avoid the US
hyperscaler budget tiers, they are optimized for a different use case.
If you want to serve real users at real volume, stop pretending a shared
VPS is the answer. Use an API, or skip straight to a GPU host.
Related reading
- AI-Powered Hosting: What Providers Actually Offer in 2026
- Managed vs Unmanaged VPS: When Is the Premium Worth It?
- Vultr vs Linode vs DigitalOcean: Developer Cloud Hosting Compared
Originally published at hostingsift.com, where we track hosting prices nightly across 50 providers.
Top comments (0)