LLMReference — Data Quality Dashboard
Pulse summarizes weekly release velocity, freshness, cadence, and current frontier pricing. Changelog is the month-by-month archive; this page covers raw field completeness for maintainers.
Headline Coverage
1,940
Total model rows tracked
Full seed catalog, including embeddings and deprecated rows
1,796
Active public models
1,796 non-deprecated language-model rows shown by default on /models
1,900
Public models incl. deprecated
104 deprecated public rows remain available behind the toggle
1,836
Active tracked rows
1,836 non-deprecated rows in the full seed, including embeddings
0
Models added this month
Released in August 2026
1,249
Open-weights models
Self-hostable or open-license rows
143
API providers covered
Provider catalogs and routes
249
Labs and creators
Research labs, companies, and builders
168
Benchmarks covered
Evaluation sources in the seed data
Database Totals
1,796
Active public models
License-Derived Openness
702
Open source
39% of active public models
500
Open weights
28% of active public models
584
Proprietary
33% of active public models
10
Unknown
1% of active public models
| License coverage | Known | Missing / unknown |
|---|
| Active public models | 1,730 (96%) | 66 (4%) |
| Families | 496 (91%) | 51 (9%) |
| OSI classification | Count | Share |
|---|
| OSI-approved effective license | 702 / 1,796 | 39% |
| Non-OSI effective license | 1,028 / 1,796 | 57% |
| Missing / unknown effective license | 66 / 1,796 | 4% |
| Commercial use | Models | Share |
|---|
| Yes | 765 / 1,796 | 43% |
| Conditional | 915 / 1,796 | 51% |
| No | 50 / 1,796 | 3% |
| Unknown / blank | 66 / 1,796 | 4% |
| Verification backlog | Count | Health |
|---|
| License records marked verify:true | 5 / 43 | 5 (12%) |
| License records with unknown openness | 1 / 43 | 1 (2%) |
| License records with unknown commercial use | 1 / 43 | 1 (2%) |
| Models using verify:true licenses | 25 / 1,796 | 25 (1%) |
| Models missing/unknown effective license | 66 / 1,796 | 66 (4%) |
| Families missing/unknown license | 51 / 547 | 51 (9%) |
Model rows use the effective license resolver: model license first, then family license. Missing licenses and licenses classified as unknown remain visible as backlog instead of being inferred from legacy strings.
Data Freshness
| Entity | Total | Placeholder 2026-01-01 | Stale >90d | Stale >180d |
|---|
| Active public models | 1,796 | 231 (13%) | 477 (27%) | 231 (13%) |
| Families | 547 | 137 (25%) | 437 (80%) | 137 (25%) |
| Providers | 143 | 33 (23%) | 38 (27%) | 33 (23%) |
| Benchmarks | 168 | 0 (0%) | 88 (52%) | 0 (0%) |
Missing or invalid lastResearched dates are counted as stale so research queues do not hide unverified rows.
Benchmark Coverage
Source URL coverage
97%
1,377 / 1,416 scores start with http.
Benchmarks with stale score sets
2
- gaokao: 11 rows, latest 2023-05-20
- gaokao-mm: 7 rows, latest 2024-02-23
| Metric | Count | Coverage / Health |
|---|
| Active public models with >=1 benchmark score | 313 / 1,796 | 17% |
| Scores with verified URL source | 1,377 / 1,416 | 97% |
| Benchmarks with zero scores | 59 / 168 | 65% |
| Benchmark scores older than 1 year | 82 / 1,416 | 94% |
| Benchmark scores older than 2 years | 18 / 1,416 | 99% |
| Benchmarks with all scores older than 2 years | 2 / 109 | 98% |
Model Field Completeness
| Group | Field | Filled | Coverage |
|---|
| Core | parameters | 1,283 / 1,796 | 71% |
| Core | context | 1,340 / 1,796 | 75% |
| Core | knowledgeCutoff | 579 / 1,796 | 32% |
| Core | description | 1,796 / 1,796 | 100% |
| Core | release | 1,775 / 1,796 | 99% |
| Core | familySlug | 1,793 / 1,796 | 100% |
| Capability | vision | 322 / 1,796 | 18% |
| Capability | reasoning | 189 / 1,796 | 11% |
| Capability | functionCalling | 276 / 1,796 | 15% |
| Capability | toolUse | 259 / 1,796 | 14% |
| Capability | structuredOutputs | 421 / 1,796 | 23% |
| Capability | codeExecution | 91 / 1,796 | 5% |
| Capability | promptCaching | 47 / 1,796 | 3% |
| Capability | batchApi | 34 / 1,796 | 2% |
| Capability | audio | 117 / 1,796 | 7% |
| Capability | fineTuning | 8 / 1,796 | 0% |
| Capability | multimodal | 424 / 1,796 | 24% |
| Metadata | maxOutputTokens | 86 / 1,796 | 5% |
| Metadata | licenseSlug | 465 / 1,796 | 26% |
Capability rows count truthy capability flags; empty, 0, and false mean the capability is not currently recorded as present.
Family and Concept Completeness
| Field | Filled | Coverage |
|---|
| family.description | 456 / 547 | 83% |
| family.researcherSlug | 545 / 547 | 100% |
| concept.lastResearched | 69 / 69 | 100% |
Pricing Coverage
| Metric | Count | Coverage / Health |
|---|
| Provider links with pricing | 1,480 / 1,789 | 83% |
| Excludes non-token-priced rows (per-image, per-second, gpu-hour, etc.). |
| Models with ≥1 pricing record | 738 / 1,796 | 41% |
| Provider links with model ID (providerModelId) | 904 / 1,995 | 45% |
| Non-token links missing all prices | 354 / 1,995 | 82% |
| Provider links with $0 token price | 24 / 1,995 | 99% |
Missing-price rows have no token, image, video, audio, query, or gpu-hour price recorded. $0 token rows may include true free-tier entries and should be reviewed before changing seed data.
Data Quality
| Indicator | Count |
|---|
| Duplicate model slugs | 0 |
| Duplicate model+provider pairs | 0 |
| Models without family | 3 |
| Families missing description | 91 |
| Concepts missing lastResearched | 0 |
| Broken family references (model→family) | 0 |
| Broken researcher references (family→researcher) | 0 |
| Broken provider references (link→provider) | 0 |
| Broken model references (link→model) | 0 |
Provider Link Completeness
| Field | Filled | Coverage |
|---|
| docs | 97 / 143 | 68% |
| pricing | 88 / 143 | 62% |
| portal | 84 / 143 | 59% |
| tier | 121 / 143 | 85% |
| providerType | 127 / 143 | 89% |
Provider Coverage (top 25 by model count)
| Provider | Models | Priced | Pricing % |
|---|
| OpenRouter | 220 | 216 | 98% |
| Fireworks AI | 211 | 207 | 98% |
| Vercel AI Gateway | 152 | 142 | 93% |
| GCP Vertex AI | 109 | 84 | 77% |
| AWS Bedrock | 108 | 95 | 88% |
| Microsoft Foundry | 106 | 79 | 75% |
| Together AI | 100 | 98 | 98% |
| Replicate API | 99 | 86 | 87% |
| Novita AI | 99 | 99 | 100% |
| DeepInfra | 52 | 49 | 94% |
| Alibaba Cloud PAI-EAS | 44 | 23 | 52% |
| Google AI Studio | 40 | 21 | 53% |
| Cloudflare Workers AI | 32 | 17 | 53% |
| OpenAI API | 31 | 31 | 100% |
| IBM watsonx | 31 | 31 | 100% |
| Volcengine | 23 | 9 | 39% |
| Mistral AI Studio | 18 | 16 | 89% |
| Bitdeer AI | 18 | 18 | 100% |
| Anthropic | 17 | 15 | 88% |
| Prodia | 16 | 0 | 0% |
| Baseten API | 14 | 1 | 7% |
| Zhipu AI GLM API | 14 | 0 | 0% |
| Databricks Foundation Model Serving | 13 | 5 | 38% |
| SiliconFlow | 13 | 13 | 100% |
| OctoAI API (Deprecated) | 12 | 0 | 0% |
Model Releases by Month
| Month | New Models |
|---|
| 2026-07 | 13 |
| 2026-06 | 51 |
| 2026-05 | 40 |
| 2026-04 | 61 |
| 2026-03 | 58 |
| 2026-02 | 39 |
| 2026-01 | 30 |
| 2025-12 | 48 |
| 2025-11 | 35 |
| 2025-10 | 32 |
| 2025-09 | 29 |
| 2025-08 | 30 |
| 2025-07 | 18 |
| 2025-06 | 18 |
| 2025-05 | 17 |
| 2025-04 | 44 |
| 2025-03 | 42 |
| 2025-02 | 17 |
| 2025-01 | 75 |
| 2024-12 | 35 |
| 2024-11 | 38 |
| 2024-10 | 43 |
| 2024-09 | 55 |
| 2024-08 | 39 |
| 2024-07 | 51 |
| 2024-06 | 77 |
| 2024-05 | 53 |
| 2024-04 | 57 |
| 2024-03 | 36 |
| 2024-02 | 63 |
| 2024-01 | 81 |
| 2023-12 | 80 |
| 2023-11 | 67 |
| 2023-10 | 45 |
| 2023-09 | 15 |
| 2023-08 | 35 |
| 2023-07 | 37 |
| 2023-06 | 23 |
| 2023-05 | 19 |
| 2023-04 | 28 |
| 2023-03 | 29 |
| 2023-02 | 2 |
| 2023-01 | 12 |
| 2022-12 | 3 |
| 2022-11 | 1 |
| 2022-10 | 12 |
| 2022-09 | 1 |
| 2022-07 | 7 |
| 2022-04 | 2 |
| 2022-03 | 2 |
| 2022-02 | 1 |
| 2022-01 | 1 |
| 2021-08 | 5 |
| 2021-05 | 3 |
| 2020-06 | 5 |
| 2020-01 | 5 |
| 2019-11 | 1 |
| 2019-08 | 4 |
| 2019-02 | 2 |
| 2018-10 | 2 |
| 2018-06 | 1 |
Generated at build time from seed JSON. Last updated: 2026-08-10.