File size: 5,191 Bytes
22f6445
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
---
pretty_name: InfraRabbit compute prices
license: cc-by-4.0
language:
- en
size_categories:
- 10K<n<100K
task_categories:
- tabular-regression
- time-series-forecasting
tags:
- pricing
- gpu
- cloud
- llm
- vps
- time-series
configs:
- config_name: indices
  data_files: "indices/*.csv"
  default: true
- config_name: snapshot
  data_files: "snapshot/*.parquet"
- config_name: history_90d
  data_files: "history/history-90d.csv"
- config_name: changes_90d
  data_files: "history/changes-90d.csv"
- config_name: all_index_series
  data_files: "history/index.csv"
- config_name: fx
  data_files: "history/fx.csv"
---

# InfraRabbit compute prices

Published list prices of compute, archived daily by [InfraRabbit](https://infrarabbit.com): VPS plans, GPU instance hours, LLM API tokens, open-model inference hosts and media models. Prices come from each provider's own price page or API. They are **list prices, not transacted prices**. A price that cannot be verified is left empty; nothing is interpolated.

Source of truth and live data: <https://infrarabbit.com/prices/data> · all indices: <https://infrarabbit.com/prices/indices> · methodology: <https://infrarabbit.com/methodology>

## Update cadence

Manual for now. This repository is a copy of the public InfraRabbit CSV downloads, refreshed by hand (a scheduled refresh may follow). The `snapshot` file name carries the export date. The site itself recomputes daily.

## Configs and columns

| Config | Files | Content |
|---|---|---|
| `indices` | `indices/<slug>.csv` | Daily index values: `date`, `value`, `n_offers` (providers or models in the basket). One file per index; slug = file name. |
| `snapshot` | `snapshot/skus-YYYY-MM-DD.parquet` (+ `.csv`) | Every active SKU with its current list price (about 21k rows). |
| `history_90d` | `history/history-90d.csv` | Every price change point observed in the last 90 days. |
| `changes_90d` | `history/changes-90d.csv` | Change feed: price changes, new SKUs, delistings. |
| `all_index_series` | `history/index.csv` | All index series (floors per GPU card, VPS floors per size, token floors, ...), all days. |
| `fx` | `history/fx.csv` | ECB USD-per-EUR rates used, per day. |

Snapshot columns: `sku_id` (category/provider/plan), `category` (vps, gpu, llm, inference, media), `provider`, `plan`, `name`, `price`, `currency`, `unit` (per_month, per_hour, per_1m_tokens_in), `price_usd`, `price_eur`, `vcpu`, `ram_gb`, `disk_gb`, `gpu`, `gpu_count`, `context`, `per_hour`, `per_month`, `per_1m_tokens_out`, `per_1m_tokens_cached_in`, `per_1m_tokens_in_batch`, `per_1m_tokens_out_batch`, `region`, `eu`, `source_id`, `source_url`, `verified_at` (last good snapshot containing the price), `price_since`, `listing_status`, `listing_first_missing`, `fx_usd_per_eur`, `fx_date`.

Per-SKU `first_seen`/`last_seen` is not in the public bulk files; `price_since`, `verified_at` and `listing_first_missing` are the closest fields. Per-SKU history beyond 90 days is not public (see the licence matrix).

## Indices (method summary)

All indices: one value per UTC day, computed from the list prices archived that day; list prices only (spot, community-cloud, marketplace and committed prices excluded); excluding VAT; USD/EUR converted at the day's ECB rate; a source not verified in time drops out; nothing is estimated.

- `h100`, `h200`, `b200`: median across providers of each provider's lowest on-demand list price per GPU-hour, in USD. Price per GPU-hour = instance hourly price / gpu_count. For H100, SXM and PCIe count, NVL excluded. Hyperscalers are included and tagged on the site.
- `eu-vps-4gb`: median across providers of each provider's cheapest monthly list price, in EUR, for a VPS with 4 up to (not including) 5 GB RAM deployable in an EU data centre; no GPU, no bare metal. Hourly plans converted at hourly x 730.
- `eu-vps-4gb-cheapest-quartile`: same selection, 25th percentile instead of the median. Minimum coverage 4 providers.
- `llm-frontier`: median blended price (3 parts input : 1 part output, USD per 1M tokens) of a frontier-model basket with first-party prices.
- `llm-open`: median blended price (3:1, USD per 1M tokens) of open-weight models, each served by at least three hosts.

`n_offers` is the number of constituents that day; it can change when a provider or model enters or leaves (see the index pages for notes).

## Licence

CC BY 4.0, as stated at <https://infrarabbit.com/data> and <https://infrarabbit.com/prices/data>: current prices, history shown, named index series and the change feed are CC BY 4.0; give credit, link, say the data date. Per-SKU history beyond 90 days and bulk history exports are a paid-plan product and are not in this dataset. Provider names, logos and trade marks belong to their owners.

Suggested attribution: `Source: InfraRabbit (infrarabbit.com/prices), data as of <date>. CC BY 4.0.`

## Citation

```bibtex
@misc{infrarabbit_compute_prices,
  author       = {InfraRabbit},
  title        = {InfraRabbit compute prices},
  year         = {2026},
  howpublished = {\url{https://infrarabbit.com}},
  note         = {List prices, not transacted prices; CC BY 4.0. Hugging Face: InfraRabbit/compute-prices}
}
```