HetznerCAX31 · 8 vCPU · 16 GB
Cost-Optimized · Arm64
Hetzner CAX31 costs $24.22/mo for 8 vCPU, 16 GB of RAM and 160 GB NVMe, cheaper than 69% of the 524 VPS plans we track, checked 16 Sept 2026.
Price
$24.22/mo
billed €20.99/mo
$0.0388/hour
$1.51 per GB RAM
Hetzner shows this plan as not available to order right now.
Cheaper than 69% of tracked offers
Specs vs the whole index
- RAM
- 16 GB> 53% of offers
- vCPU
- 8 shared> 71% of offers
- Disk
- 160 GB NVMe> 66% of offers
- Bandwidth
- Not published
Excl. VAT · IPv4 +€0.50/mo extra · priced at Nuremberg, DE
Value score
89/100
- RAM · 16 GB
- 16.0
- vCPU · 8
- 12.0
- Disk · 160 GB NVMe
- 6.4
- Traffic · not published, median 6.1 TB
- 1.8
- 36.2 points ÷ $24.22
- 1.50 per $
1 point per GB of RAM, 1.5 per vCPU, 1 per 25 GB of disk up to 500 GB (SSD × 0.8, HDD × 0.4), and 0.3 per TB of traffic up to 20 TB.
The score ranks 1.50 points per dollar against the 524 tracked VPS plans, whose median is 0.43. A higher score means more machine for the money.
Regions · 3
- Helsinki, FI
- Nuremberg, DE
- Falkenstein, DE
Data provenance
Hetzner · Live web · hetzner.com cloud plan pages + live_data_prices.json price feed
Fetched 16 Sept 2026
Prices exclude VAT. Always confirm on the provider's site.
One-click via Hetzner Cloud Apps: Docker · WordPress · Coolify · Nextcloud · WireGuard (+12-app catalog) matrix →
Where this sits in the market
At $24.22/mo, CAX31 · 8 vCPU · 16 GB is cheaper than 69% of the 524 VPS plans in our index. It delivers 16 GB of RAM at $1.51/GB, 72% below the $5.42/GB market median. Its value score of 89/100 ranks it #53 among VPS plans that include a disk, and it is the 9th cheapest of Hetzner's 22 tracked plans.
Open models this box runs comfortably · CPU inference
how we compute this →- Qwen3 14B~2.1–4.3 tok/s
- DeepSeek-R1 Distill Qwen 14B~2.1–4.3 tok/s
- Phi-4 14B~2.2–4.3 tok/s
- Mistral Nemo 12B~2.6–5.2 tok/s
- Gemma 3 12B~2.6–5.2 tok/s
- Qwen3 8B~3.9–7.7 tok/s
- DeepSeek-R1-0528 Qwen3 8B~3.9–7.7 tok/s
- Llama 3.1 8B~3.9–7.9 tok/s
Largest two per lineup that fit 16 GB with 20% headroom at Q4_K_M and an 8K context; speeds are bandwidth-bound estimates for 8 shared vCPU.
