Compare what cloud, GPU and LLM providers actually charge.
184 providers · 701 models across 1750 per-token price listings · 267 GPU rates · 144 hosting plans. Every number links to its source, last collected 2026-09-18.
LLM APIs & routers
- Model labs 24
First-party APIs from the companies that train the models. - Inference hosts 31
Serve open-weights models on their own GPUs, priced per token. - Hyperscaler AI platforms 5
Bedrock, Vertex AI, Azure AI Foundry: many models, one cloud bill. - LLM routers 14
One API key, many upstream providers. They resell tokens. - LLM gateways 12
Proxy, fallback and observability over your own provider keys.
GPU compute
- GPU clouds 29
Rent GPUs by the hour, from hyperscalers to peer marketplaces.
Cloud hosting
- Hyperscalers 7
General-purpose clouds with every service and every region. - Cloud VPS 21
Virtual machines you manage yourself, billed by plan. - App platforms (PaaS) 20
Push code or a container; the platform runs and scales it. - Serverless containers 5
Request-billed containers that scale to zero. - Frontend & static hosting 5
Static sites and frontend frameworks with a CDN attached. - Edge compute 5
Code that runs in CDN points of presence. - Managed Kubernetes 9
Hosted control planes; you pay for the nodes. - Object storage 7
S3-compatible storage, where egress is the real price. - Self-hosted control planes 7
Run your own PaaS on a VPS you rent. - Managed hosting 3
Managed WordPress and app hosting on someone else's cloud.
Most widely served models
All models →| Model | Providers | Cheapest input /1M | Cheapest output /1M | Spread |
|---|---|---|---|---|
| gpt-oss-120b openai/gpt-oss-120b |
30 | $0.03 | $0.14 | 7.6× |
| Llama 3.3 70B Instruct meta-llama/llama-3.3-70b-instruct |
22 | $0.10 | $0.32 | 7.9× |
| GLM-5.2 z-ai/glm-5.2 |
22 | $0.561 | $1.76 | 3.6× |
| GLM-5.3 z-ai/glm-5.3 |
22 | $0.90 | $3 | 1.9× |
| Kimi K2.6 moonshotai/kimi-k2.6 |
21 | $0.50 | $2.85 | 2.1× |
| gpt-oss-20b openai/gpt-oss-20b |
19 | $0.02 | $0.10 | 5.6× |
| GLM-5.3 Flash z-ai/glm-5.3-flash |
19 | $0.09 | $0.28 | 1.7× |
| DeepSeek V4 Flash 0731 deepseek/deepseek-v4-flash-0731 |
18 | $0.06 | $0.12 | 10.6× |
| Kimi K3 moonshotai/kimi-k3 |
18 | $2.10 | $10.95 | 1.7× |
| Gemma 4 31B google/gemma-4-31b-it |
17 | $0 | $0 | |
| DeepSeek V3.2 deepseek/deepseek-v3.2 |
17 | $0.25 | $0.38 | 11.6× |
| Qwen3.8-27B qwen/qwen3.8-27b |
16 | $0.20 | $1.49 | 2.2× |
Spread is the most expensive provider's blended price divided by the cheapest's, for the same model.