datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
gspc-hf-model-census
GSPC — Hugging Face model census
20,000 models on the Hub, and this file measures none of them. That is the point of it: it
is the denominator, not the score. Every row reads status: UNMEASURED, measured_axes: 0,
unmeasured_axes: 22.
What changed in 0.2, and why it mattered
The 0.1 census carried 15,557 rows, each asserting unmeasured_axes: 14. Fourteen was the
superseded 13-of-14 canon. The board has been ruled at 22 axes and has read
22 slots · 22 measured · 0… See the full description on the dataset page: https://huggingface.co/datasets/csoai/gspc-hf-model-census.claim-capture-census
Claim-capture census
A daily capture of the claims that public, authless catalogues serve about themselves and about
the things they list. Produced by census-capture.py on CSOAI infrastructure.
A capture is CLAIM_CAPTURED. It is not a measurement, not a grade, and not a certification.
The only thing a capture proves is that these records existed in this exact form at this time as
served by that source. Every artifact carries that boundary in claim_boundary.
What is… See the full description on the dataset page: https://huggingface.co/datasets/csoai/claim-capture-census.oss-model-census
Open-weight model census
DISCOVERED — every row in this dataset is a census listing, not a measurement. n_measured = 0. Live board: https://councilof.ai/api/gspc
A frozen snapshot of open-licence text-generation models on the Hugging Face Hub.
n_frozen: 68,869 models (from 120,000 raw listings; slice: pipeline_tag=text-generation, sorted by downloads desc, first 120 pages of 1,000, kept only models with an explicit open-licence tag).
as_of: 2026-09-01T04:41:36Z (UTC), crawled… See the full description on the dataset page: https://huggingface.co/datasets/csoai/oss-model-census.agent-interop-census
Agent Interop Census
A census, not a scoreboard. 7,645 rows indexing what publicly exists in the
agent-interoperability ecosystem: MCP servers in the official registry, Hugging Face Spaces,
and open-licence repositories on GitHub.
Zero rows carry a measurement. graded: false and measurement: null on every single row.
A grade requires a run against a frozen bank; no run was performed here. Each row states, in
the row itself, what it is (an index entry observed in a public… See the full description on the dataset page: https://huggingface.co/datasets/csoai/agent-interop-census.x402-settlement-census
x402 settlement census — 2026-09-06
We paid 316 conformant x402 hosts as an ordinary buyer and recorded what came back.
Not a survey of what hosts advertise — a record of what they did when real USDC arrived.
outcome
hosts
share
REFUSED
213
67.4%
DELIVERED
100
31.6%
NO_CHALLENGE
2
0.6%
MISMATCH
1
0.3%
Two in three conformant hosts refused a correctly-signed payment. Being listed in a Bazaar index
and answering a valid 402 is not the same as taking money and… See the full description on the dataset page: https://huggingface.co/datasets/csoai/x402-settlement-census.fitllm-fit-census
Local LLM Fit Census v1 — 2026-09-13
10,530 verdicts: 30 models × 93 devices (36 GPUs + 57 Mac configs) × per-platform quant tiers.
Each row is generated by fitllm-engine from architecture inputs pinned to official config.json files. Runtime and OS reserves remain documented estimates. Reproduce it yourself: npm run census.
Assumptions: context = min(8K, model max) · KV cache F16 · platform reserve/headroom per engine. Interactive per-combo pages: fitllm.run/can-i-run.… See the full description on the dataset page: https://huggingface.co/datasets/click6067/fitllm-fit-census.erc8004-base-census-jun2026
ERC-8004 Base Mainnet Census — June 2026
Read-only census of AI agents registered under ERC-8004 (Trustless Agents) on Base mainnet, taken at block 47,041,190 (June 7, 2026).
Headline numbers (full scan + uniform sample):
54,802 agents registered in the IdentityRegistry (0x8004A1...a432, deployed Feb 3, 2026)
Uniform random sample of 2,000 agents queried against the ReputationRegistry (0x8004BAa1...9b63):
52.8% have at least one feedback client; median 1 client per agent, max… See the full description on the dataset page: https://huggingface.co/datasets/rsoft-latam/erc8004-base-census-jun2026.