Dataset Viewer
The dataset viewer is not available for this split.
Cannot load the dataset split (in streaming mode) to extract the first rows.
Error code: StreamingRowsError
Exception: CastError
Message: Couldn't cast
$schema: string
title: string
description: string
type: string
items: struct<type: string, required: list<item: string>, properties: struct<provider: struct<type: list<it (... 1003 chars omitted)
child 0, type: string
child 1, required: list<item: string>
child 0, item: string
child 2, properties: struct<provider: struct<type: list<item: string>, description: string>, instance_type: struct<type: (... 939 chars omitted)
child 0, provider: struct<type: list<item: string>, description: string>
child 0, type: list<item: string>
child 0, item: string
child 1, description: string
child 1, instance_type: struct<type: list<item: string>, description: string>
child 0, type: list<item: string>
child 0, item: string
child 1, description: string
child 2, gpu_type: struct<type: list<item: string>, description: string>
child 0, type: list<item: string>
child 0, item: string
child 1, description: string
child 3, num_gpus: struct<type: list<item: string>, description: string>
child 0, type: list<item: string>
child 0, item: string
child 1, description: string
child 4, region: struct<type: list<item: string>, description: string>
child 0, type: list<item: string>
child 0, item: string
child 1, description: string
child 5, region_name: struct<type: list<item: string>, descriptio
...
type (... 159 chars omitted)
child 0, @type: string
child 1, @id: string
child 2, name: string
child 3, description: string
child 4, field: list<item: struct<@type: string, @id: string, name: string, description: string, dataType: string, s (... 81 chars omitted)
child 0, item: struct<@type: string, @id: string, name: string, description: string, dataType: string, source: stru (... 69 chars omitted)
child 0, @type: string
child 1, @id: string
child 2, name: string
child 3, description: string
child 4, dataType: string
child 5, source: struct<fileObject: struct<@id: string>, extract: struct<column: string>>
child 0, fileObject: struct<@id: string>
child 0, @id: string
child 1, extract: struct<column: string>
child 0, column: string
creator: struct<@type: string, name: string, url: string>
child 0, @type: string
child 1, name: string
child 2, url: string
license: string
datePublished: timestamp[s]
keywords: list<item: string>
child 0, item: string
conformsTo: string
@type: string
@context: struct<@vocab: string, cr: string, sc: string, data: struct<@id: string, @type: string>>
child 0, @vocab: string
child 1, cr: string
child 2, sc: string
child 3, data: struct<@id: string, @type: string>
child 0, @id: string
child 1, @type: string
identifier: string
to
{'@context': {'@vocab': Value('string'), 'cr': Value('string'), 'sc': Value('string'), 'data': {'@id': Value('string'), '@type': Value('string')}}, '@type': Value('string'), 'conformsTo': Value('string'), 'name': Value('string'), 'description': Value('string'), 'url': Value('string'), 'license': Value('string'), 'creator': {'@type': Value('string'), 'name': Value('string'), 'url': Value('string')}, 'datePublished': Value('timestamp[s]'), 'version': Value('string'), 'identifier': Value('string'), 'keywords': List(Value('string')), 'temporalCoverage': Value('timestamp[s]'), 'distribution': List({'@type': Value('string'), '@id': Value('string'), 'name': Value('string'), 'contentUrl': Value('string'), 'encodingFormat': Value('string'), 'sha256': Value('string')}), 'recordSet': List({'@type': Value('string'), '@id': Value('string'), 'name': Value('string'), 'description': Value('string'), 'field': List({'@type': Value('string'), '@id': Value('string'), 'name': Value('string'), 'description': Value('string'), 'dataType': Value('string'), 'source': {'fileObject': {'@id': Value('string')}, 'extract': {'column': Value('string')}}})})}
because column names don't match
Traceback: Traceback (most recent call last):
File "/src/services/worker/src/worker/utils.py", line 147, in get_rows_or_raise
return get_rows(
dataset=dataset,
...<4 lines>...
column_names=column_names,
)
File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
return func(*args, **kwargs)
File "/src/services/worker/src/worker/utils.py", line 127, in get_rows
rows_plus_one = list(itertools.islice(safe_iter(ds, dataset=dataset), rows_max_number + 1))
File "/src/services/worker/src/worker/utils.py", line 483, in safe_iter
yield from ds.decode(False) if ds.features else ds
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2840, in __iter__
for key, example in ex_iterable:
^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2373, in __iter__
for key, pa_table in self._iter_arrow():
~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2398, in _iter_arrow
for key, pa_table in self.ex_iterable._iter_arrow():
~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 536, in _iter_arrow
for key, pa_table in iterator:
^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 419, in _iter_arrow
for key, pa_table in self.generate_tables_fn(**gen_kwags):
~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 343, in _generate_tables
self._cast_table(pa_table, json_field_paths=json_field_paths),
~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
pa_table = table_cast(pa_table, self.info.features.arrow_schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2378, in table_cast
return cast_table_to_schema(table, schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2306, in cast_table_to_schema
raise CastError(
...<3 lines>...
)
datasets.table.CastError: Couldn't cast
$schema: string
title: string
description: string
type: string
items: struct<type: string, required: list<item: string>, properties: struct<provider: struct<type: list<it (... 1003 chars omitted)
child 0, type: string
child 1, required: list<item: string>
child 0, item: string
child 2, properties: struct<provider: struct<type: list<item: string>, description: string>, instance_type: struct<type: (... 939 chars omitted)
child 0, provider: struct<type: list<item: string>, description: string>
child 0, type: list<item: string>
child 0, item: string
child 1, description: string
child 1, instance_type: struct<type: list<item: string>, description: string>
child 0, type: list<item: string>
child 0, item: string
child 1, description: string
child 2, gpu_type: struct<type: list<item: string>, description: string>
child 0, type: list<item: string>
child 0, item: string
child 1, description: string
child 3, num_gpus: struct<type: list<item: string>, description: string>
child 0, type: list<item: string>
child 0, item: string
child 1, description: string
child 4, region: struct<type: list<item: string>, description: string>
child 0, type: list<item: string>
child 0, item: string
child 1, description: string
child 5, region_name: struct<type: list<item: string>, descriptio
...
type (... 159 chars omitted)
child 0, @type: string
child 1, @id: string
child 2, name: string
child 3, description: string
child 4, field: list<item: struct<@type: string, @id: string, name: string, description: string, dataType: string, s (... 81 chars omitted)
child 0, item: struct<@type: string, @id: string, name: string, description: string, dataType: string, source: stru (... 69 chars omitted)
child 0, @type: string
child 1, @id: string
child 2, name: string
child 3, description: string
child 4, dataType: string
child 5, source: struct<fileObject: struct<@id: string>, extract: struct<column: string>>
child 0, fileObject: struct<@id: string>
child 0, @id: string
child 1, extract: struct<column: string>
child 0, column: string
creator: struct<@type: string, name: string, url: string>
child 0, @type: string
child 1, name: string
child 2, url: string
license: string
datePublished: timestamp[s]
keywords: list<item: string>
child 0, item: string
conformsTo: string
@type: string
@context: struct<@vocab: string, cr: string, sc: string, data: struct<@id: string, @type: string>>
child 0, @vocab: string
child 1, cr: string
child 2, sc: string
child 3, data: struct<@id: string, @type: string>
child 0, @id: string
child 1, @type: string
identifier: string
to
{'@context': {'@vocab': Value('string'), 'cr': Value('string'), 'sc': Value('string'), 'data': {'@id': Value('string'), '@type': Value('string')}}, '@type': Value('string'), 'conformsTo': Value('string'), 'name': Value('string'), 'description': Value('string'), 'url': Value('string'), 'license': Value('string'), 'creator': {'@type': Value('string'), 'name': Value('string'), 'url': Value('string')}, 'datePublished': Value('timestamp[s]'), 'version': Value('string'), 'identifier': Value('string'), 'keywords': List(Value('string')), 'temporalCoverage': Value('timestamp[s]'), 'distribution': List({'@type': Value('string'), '@id': Value('string'), 'name': Value('string'), 'contentUrl': Value('string'), 'encodingFormat': Value('string'), 'sha256': Value('string')}), 'recordSet': List({'@type': Value('string'), '@id': Value('string'), 'name': Value('string'), 'description': Value('string'), 'field': List({'@type': Value('string'), '@id': Value('string'), 'name': Value('string'), 'description': Value('string'), 'dataType': Value('string'), 'source': {'fileObject': {'@id': Value('string')}, 'extract': {'column': Value('string')}}})})}
because column names don't matchNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
Scrutica Compute Cost Index
Cloud GPU rental prices across on-demand, spot and reserved tiers, with costs per petaFLOP-hour, day and year calculated from peak dense BF16 throughput. The normalization excludes structured sparsity and does not adjust for achieved workload utilization.
Scrutica Compute Cost Index — v2026.3
Release date: 2026-07-17
Documentation revised: 2026-09-08
License: CC-BY-4.0
Suggested citation: see CITATION block at the bottom.
Coverage
508 instance and pricing-tier observations across 6 providers and 9 accelerator types. Pricing tiers are on-demand, spot, one-year reserved and three-year reserved.
Normalisation
The calculation divides the hourly instance price by its peak dense BF16 throughput in PFLOP/s. Instance throughput is the GPU count multiplied by per-GPU TFLOP/s and divided by 1,000. Multiplying the resulting hourly cost by 24 gives cost per petaFLOP-day; multiplying the day value by 365 gives the year value.
Throughput specifications come from Epoch AI’s ML Hardware data. The calculation excludes structured sparsity and applies no utilization discount, so it prices peak capacity. Reserved prices are expressed as hourly rates for the stated commitment term. Missing inputs leave the normalized cost blank.
Sources and authority tiers
| Source | Authority tier | What it contributes |
|---|---|---|
| AWS Pricing Bulk API | 1 | AWS instance rates |
| Azure Retail Prices API | 1 | Azure instance rates |
| GCP Compute Engine GPU pricing | 1 | Google Cloud instance rates |
| Oracle OCI price list API | 1 | OCI instance rates |
| Lambda pricing page | 1 | Lambda instance rates |
| CoreWeave pricing page | 1 | CoreWeave instance rates |
| ec2.shop JSON API | 2 | Third-party AWS price observations; on-demand prices from this source are blank in the bundle |
| gcloud-compute.com pricing tables | 2 | Third-party Google Cloud price observations |
| Epoch AI ML Hardware | 2 | Compiled accelerator specifications for the dense BF16 throughput denominator |
Refresh path
scripts/package-datasets.ts packages the observations and calculated cost fields in data/processed/cloud_pricing.json. The file’s retrieved_at column dates each observation; the release date identifies the packaged version.
Known limitations
- Spot rates are dated observations and can vary within a day.
- The files cover published rental rates. Negotiated enterprise discounts and the costs of buying and operating hardware require separate data.
- A throughput-normalized price does not account for workload-specific utilization, memory or interconnect requirements.
Citation
@dataset{scrutica_compute_cost_index_2026_3,
author = {Gringras, David and {Scrutica}},
title = {Scrutica Compute Cost Index},
year = {2026},
version = {2026.3},
publisher = {Scrutica},
note = {No DOI assigned; cite the version and SHA-256 checksum at /datasets/compute-cost-index},
url = {https://scrutica.com/datasets/compute-cost-index}
}
- Downloads last month
- 67