Skip to content
Longterm Wiki

Foundation Model Transparency Index

by Stanford CRFM·Latest: v1.2 December 2025·Source ↗

Transparency-focused index scoring developers on 100 indicators across upstream resources, the model itself, and downstream use.

Organizations
13
Dimensions
123
Snapshots
3
Latest wave
v1.2 December 2025

Grades by dimension — v1.2 December 2025

Rows are organizations; columns are the dimensions FMTI scores. Cell text is the published grade (overall first, then per-pillar).

OrganizationOverallAcceptable use policyAccountabilityAgent ProtocolsAI bug bountyAmount of usageAUP enforcement frequencyAUP enforcement processBasic model propertiesBenchmarked inferenceBenefits AssessmentCapabilitiesCapabilities evaluationCapabilities taxonomyCarbon emissions for final training runChange logClassification of usage dataCode accessComputeCompute hardware for final training runCompute providerCompute usage for final training runCompute usage including R&DConsumer/enterprise usageCrawlingData AcquisitionData acquisition methodsData domain compositionData laborer practicesData language compositionData ProcessingData processing methodsData processing purposeData processing techniquesData PropertiesData replicabilityData retention and deletion policyData sizeDeeper model propertiesDetection of machine-generated contentDevelopment duration for final training runDistribution channels with usage dataDocumentation for responsible useDownstreamDownstream mitigationsEnergy usage for final training runEnterprise mitigationsEnterprise usersExternal data accessExternal developer mitigationsExternal products and servicesExternal reproducibility of capabilities evaluationExternal reproducibility of mitigations evaluationExternal reproducibility of risks evaluationExternal risk evaluationFeedback mechanismsFoundation model roadmapGeographic statisticsGovernment commitmentsGovernment useImpactInstructions for data generationIntermediate tokensInternal compute allocationInternal product and service mitigationsInternal products and servicesLicensed data compensationLicensed data sourcesMethodsMisuse incident reporting protocolMitigations efficacyMitigations taxonomyMitigations taxonomy mapped to risk taxonomyModelModel accessModel Behavior PolicyModel costModel dependenciesModel informationModel MitigationsModel objectivesModel response characteristicsModel stagesModel theft prevention measuresNew human-generated data sourcesNotice of usage data used in trainingOpen weightsOrganization chartOther resourcesOversight mechanismPermitted and prohibited usersPermitted, restricted, and prohibited model behaviorsPermitted, restricted, and prohibited usesPost-deployment coordination with governmentPost-deployment monitoringPre-deployment risk evaluationPublic datasetsQuantizationRegional policy variationsReleaseRelease stagesResearcher creditsResponsible disclosure policyRisk thresholdsRisksRisks evaluationRisks taxonomySafe harborSecurity incident reporting protocolSpecialized accessSynthetic data purposeSynthetic data sourcesSystem promptTerms of useTop distribution channelsTrain-test overlapUpstreamUsage dataUsage data used in trainingUsers of internal products and servicesVersioning protocolWater usage for final training runWhistleblower protection
AI21 Labs66801001001000010010001007510010001000022.21001000010010091.71000100066.710010000010001001000010075100010010001000100001000100010010071.41001000100100100066.7100010010070501000100756010010010010010010010010050100100100100071.4010010010087.510001001006010010010010001001001001000052.92010001000100
Alibaba266033.3100000100100005010010001000011.10100000016.7000000002000100100000022.24001000010000000010001000001000000066.700004050500100750100100100000100000100010000000062.500000000001001000100100017.600010000
Amazon398033.3100100001000005010010001000011.10100000016.710000033.310000000001000100100501000100001000000100010000000100010000066.700100100505075000601001001001000000010010010010010057.1000100751000100100400100010010010000100100017.62000000
Anthropic4610010010000100100001002510000100100000000010025100000000000000100010010072.2100010000100000010000100100028.601000100100000010010010056.7501000100258001000100010000010010010010010057.1000100751001001001006010010010010000010010010008.860001000100
DeepSeek326033.3100000100100005010010001000044.4100100000016.7000066.71001000200010010001000019.40000000000001000100000100100000066.700100046.750750100752010010010000010000010001000000100062.50000200100000100100100100100032.4000000
Google4180100100100001000010025100001000011.1010000010033.3000033.301000000001000010055.6100010000100100000001000100028.60100010000066.700100043.35075000401001001001000100000100100100100042.900010087.5100100100100200100010001001000100100023.50001000100
IBM958010010010010001001001001007510010010010010010010010010010010010010010010010010010010010010010010010010010010010010010010091.710010010001001001001000100100100100010010085.710010010010010010010010010010010010093.310010010010010080100100100100100100100100100100100100100100100100100100100100100100100100100100100100100100100100100100100010080100100100100100
Meta3140001000001000100501001001001000022.2010000010033.3100000000020001001001000010033.380010000100000000000014.301000000000000405075010075001000001001000001000100028.6001000500100100100200100000001001000020.601000000
Midjourney14600000010000000001000000000000000000000000000100100254000001001000000010000028.60000100000000016.7025000001000000000010001000000100062.50000000000000100100002000000
Mistral186033.3100100001000002510000100000000000000000000000001000010033.3800100001000000001000100014.30000100000000020250000200001000000001000100042.9000037.500100000010000000100000000000
OpenAI356010001000000002510000100000000001008.30000000000100001000010058.31000100001000000100000100014.301000100100000010010010043.307500080010001000000010010010010010071.400010062.51000100100601001001001000000100002.920001000100
Writer728066.71001001000100100100100501001001001001000100100100100100100058.31000100100100100100100400100100010010010010083.3100100100100010010000001000100100085.71001001001001000033.3100010006050501001007540010010010001000100100100100010010085.7100010010087.5100100100100400100100001001000100100070.61001001001001000
xAI14600000010000025100001000011.10100000000000000000000000022.24001000000000001000000010001000000000016.7075000001000000000010001000000005000000000000010010010002.9000000

Snapshot history

WavePublishedOrgsLicenseLinks
v1.2 December 2025latest2025-12-0113CC-BY-4.0
v1.1 May 20242024-05-2114CC-BY-4.0
v1.0 October 20232023-10-1810CC-BY-4.0

Methodology

Transparency-focused index scoring developers on 100 indicators across upstream resources, the model itself, and downstream use.

Source ↗Methodology ↗License: CC-BY-4.0Status: active

Grades are mirrored from upstream sources. We do not score organizations ourselves — see the source link for full methodology and per-indicator detail. Citation of published grades is consistent with fair-use attribution.