Compute prices, compared.
Cloud VMs, AI APIs and hardware, pulled from official sources and dated.
Cloud virtual machines
Compare VM compute and list prices
Lowest on-demand price for 5 of 14 providers, as of . ≈ marks an estimate: the hourly rate × 730 hours.
Start with a requirement: compare VMs with 32 GiB RAM for memory capacity, VMs under $10 per month for a base-compute budget, or Hetzner and DigitalOcean at matched memory sizes for a provider shortlist. Check storage, traffic and tax separately.
| Item | Value | Detail |
|---|---|---|
| Oracle Cloud Infrastructure | ≈ $16.06 | VM.Standard.A2.Flex (1 OCPU, 4 GB) · 4 GiB · Oracle public price list; deployment region not specified |
| Vultr | $20.00 | vc2-2c-4gb · 4 GiB · Amsterdam (ams) |
| Akamai Connected Cloud | $24.00 | Linode 4GB · 4 GiB · Tokyo 2, JP |
| DigitalOcean | $24.00 | Basic (s-2vcpu-4gb) · 4 GiB · Amsterdam 3 |
| AWS | ≈ $24.53 | t4g.medium · 4 GiB · US West (Oregon) |
| Microsoft Azure | ≈ $29.78 | Standard_D2pls_v6 · 4 GiB · IN Central |
| Google Cloud | ≈ $43.75 | c2d-standard-2 · 8 GiB · asia-south1 |
Shared, dedicated and Arm CPUs are mixed: price only, not performance.
| Item | Value | Detail |
|---|---|---|
| Hetzner Cloud | €5.99 | CX23 · 4 GiB · Falkenstein (FSN1) |
| UpCloud | €12.00 | STARTER-2xCPU-4GB · 4 GiB · Sydney #1 (au-syd1) |
| Exoscale | ≈ €34.06 | Standard Medium · 4 GiB · Exoscale public price list (all zones, per the pricing page) |
Not like for like on tax: 2 providers exclude tax and 1 does not say.
Guides and providers
AI
API services and models you can run yourself
Prices per 1M tokens, as of –.
Of 19 models sold by more than one API host, 2 keep a listed input price gap of 2× or more even without the lowest host, and 8 with every host counted. See the price gaps.
24 published price changes since Sep 28, 2026. See what changed in compute prices.
| Category | Lowest value | Lowest-priced | Highest value | Items |
|---|---|---|---|---|
| DeepInfra | $0.03 | gpt-oss-20b | $2.85 | 20 |
| Amazon Bedrock | $0.035 | Nova Micro | $4 | 36 |
| SiliconFlow | $0.04 | gpt-oss-20b | $2.70 | 27 |
| Fireworks AI | $0.05 | NVIDIA Nemotron 3.5 Lightning 30B A3B | $4.50 | 19 |
| Nebius Token Factory | $0.06 | Nemotron-3.5-Lightning | $3 | 19 |
| Anthropic | $0.10 | Claude Haiku 5.5 (≤100K tokens) | $10 | 9 |
| Baseten | $0.10 | GPT OSS 120B | $3 | 18 |
| $0.10 | Gemini 2.5 Flash-Lite (text/image/video) | $2 | 4 | |
| Mistral AI | $0.10 | Ministral 3 3B | $0.70 | 6 |
| OpenAI | $0.10 | GPT-6 Luna (short context) | $10 | 9 |
| OpenRouter | $0.10 | Meta: Llama 3.3 70B Instruct (routed listing) | $0.25 | 2 |
| StepFun | $0.10 | Step-3.5-Flash | $1 | 3 |
| Novita AI | $0.135 | Llama 3.3 70B Instruct (Serverless tier) | $1.40 | 12 |
| Xiaomi MiMo | $0.14 | mimo-v2.6-flash | $4.35 | 3 |
| Alibaba Cloud Model Studio | $0.15 | Qwen3.8-Flash (≤1M tokens) | $2 | 3 |
| MiniMax | $0.15 | MiniMax-M3.1-Flash | $0.30 | 2 |
| Z.ai | $0.15 | GLM-5.3-Flash | $1.40 | 4 |
| DeepSeek | $0.30 | DeepSeek V4.1 Flash 1017 | $1.32 | 2 |
| xAI | $1 | grok-build-0.1 (<200K tokens) | $2 | 8 |
| Moonshot AI | $2 | Kimi K3.1 Pro | $4 | 3 |
14 offers without a single USD text-token input price are not plotted. 8 plotted prices are promotional or on sale.