Price checker
What GPUs and AI models cost right now, read from each provider's own price list and dated.
Each page says when its prices were read. The GPU page also shows how each price has moved over 30 days, 90 days, 180 days and 12 months.
GPU rental prices
The on-demand price per GPU-hour of 15 GPU models at 7 providers, November 2026.
| GPU | Per GPU-hour | 12 months |
|---|---|---|
| NVIDIA H100 | $4.14 | up 52.4 percent over 12 months |
| NVIDIA H200 | $5.40 | up 50.4 percent over 12 months |
| NVIDIA B200 | $8.24 | up 47.4 percent over 12 months |
| NVIDIA B300 | $8.99 to $9.50 | no data |
LLM price index
The API price per million tokens of 54 models from 8 providers: input, output, cached input and batch.
- Sort every model by input, output or blended price.
- Open a model for worked costs at real volumes and its closest rivals.
- Cheapest current model by blended rate: Amazon Nova Micro, $0.061 per 1M tokens.
AI token usage by model
Not a price: the demand behind the prices. OpenRouter processed 834 trillion AI tokens in the 30 days to 13 Nov 2026, across 589 models.
- 1. DeepSeek V4.1 Flash 101722.2% of tokens, 185T in the 30 days to 13 Nov 2026
- 2. DeepSeek V4.1 Flash9.5% of tokens, 79.4T in the 30 days to 13 Nov 2026
- 3. GLM 5.3 Flash6.9% of tokens, 57.4T in the 30 days to 13 Nov 2026
What a price here means
- A list price: what the provider publishes, before any discount you have negotiated.
- Dated: every page shows the day its prices were read, and nothing is carried forward silently.
- Sourced: each price links to the provider page or price feed it came from.
To turn a price into a monthly cost for your own workload, use the GPU Cost Calculator or the LLM Cost Calculator. The H100 has its own page: NVIDIA H100 rental price.
Want your own GPU and AI prices checked?
Cloud Cost Audit: We review your GPU machines, schedules and disks against what your models need. From $1,500, fixed. Nothing paid upfront.