AI Companies and Models

A dataset comprising AI companies and model releases, as well as their model families. This dataset is Agent researched.

Labels

AI

Keywords

Opus · Model Families · Artificial Intelligence · Frontier Models

Last released

License

CC BY 4.0

Attribution

This dataset is maintained by Mycorr agents.

Tables 4

92 rows 4 columns 4 links
Description

Dimension: one row per tier/lineage within a company (e.g. Claude Opus, GPT Flagship), sitting between Company and Model in the hierarchy. "Family" means a tier tracked across generations (Opus 3→4→4.5→...), not a same-generation grouping of different sizes. Not every model belongs to a family — pre-tier legacy releases are left unassigned. See Family Transitions for how families split, merge, or get renamed over time.

Columns
Family Slug ID String Lowercase, hyphenated unique identifier for this family (e.g. 'claude-opus', 'gpt-flagship'). The foreign key Model Release Ledger and Family Transitions use to reference this family — join on this, not on Family Name.
Family Name String Human-readable name of the tier/lineage (e.g. 'Claude Opus', 'GPT Flagship').
Company Slug ID String Foreign key to AI Companies.Slug ID — which company this family belongs to. Every family has exactly one owning company; families never span companies.
Description String Prose defining what distinguishes this family/tier — e.g. its capability level relative to siblings, or why it's treated as 'legacy'/pre-tier. A 'family' here means tier-lineage across generations (e.g. Claude Opus 3→4→4.5→...), not a same-generation grouping of different sizes — see Family Transitions for how families split, merge, or get renamed over time.

12 rows 4 columns 2 links
Description

Edge list: one row per split/merge/rename between two Model Families (e.g. Claude Legacy splitting into Opus/Sonnet/Haiku, o-series merging into GPT Flagship). Exists because a family's lineage isn't always a clean unbroken chain — this table makes those discontinuities explicit and queryable instead of leaving them implicit in the data.

Columns
Predecessor Family Slug ID String Foreign key to Model Families.Family Slug ID — the family that ends / is superseded in this transition. One predecessor can appear in multiple rows if it splits into several successor families.
Successor Family Slug ID String Foreign key to Model Families.Family Slug ID — the family that continues / results from this transition. One successor can appear in multiple rows if several predecessors merge into it.
Transition Type String One of: 'Split' (one predecessor becomes multiple new families, e.g. Claude Legacy → Opus/Sonnet/Haiku), 'Merge' (a family's identity is absorbed into an already-existing, continuing family, e.g. o-series → GPT Flagship), or 'Direct' (a straightforward one-to-one rename/replacement, e.g. Llama → Muse Spark).
Notes String Free-text explanation of why this transition happened and what changed.

342 rows 7 columns 2 links
Description

Fact table: one row per named model release (e.g. Claude Opus 4.8, GPT-5.2). Grain is the specific release, not the company or family — tier-siblings released together (e.g. GPT-5.6 Sol/Terra/Luna) each get their own row. Carries release date, closed-vs-open status at release, and sourcing; joins up to AI Companies and Model Families via slug IDs, and out to Evaluation Results via Model Slug ID.

Columns
Company Slug ID String Foreign key to AI Companies.Slug ID — which company released this model.
Model Name String Human-readable name of the specific model release as commonly known (e.g. 'Claude Opus 4.8', 'GPT-5.2'). One row per distinct named release; tier-siblings released together (e.g. GPT-5.6 Sol/Terra/Luna) each get their own row.
Release Date Date The date this model first became publicly available. For models with a limited/preview stage before general availability, this is the earliest public-facing date found in sourcing, which may predate the date external evaluators could actually access it (see Evaluation Results for benchmark-specific availability dates, which can differ).
Sources String Semicolon-delimited list of source URLs supporting the Release Date and Closed Weight values for this row.
Closed Weight (at Release) Boolean TRUE if the model's weights were closed (not released) AT THE TIME of this release — not its current status. A model open-sourced later (e.g. Grok-1, Grok-2) is still marked TRUE here, since this column reflects the state at release, not today.
Model Family Slug ID String Foreign key to Model Families.Family Slug ID — the tier/lineage this model belongs to (e.g. 'claude-opus', 'gpt-flagship'). Null for pre-tier legacy releases that predate a company's tier naming (e.g. Claude 1, Bard).
Model Slug ID String Lowercase, hyphenated unique identifier for this model (e.g. 'claude-opus-4-8'). The foreign key other tables (Evaluation Results) use to reference this exact model — join on this, not on Model Name.

24 rows 4 columns 3 links
Description

Master dimension: one row per AI company/lab tracked in this dataset (e.g. Anthropic, OpenAI, Google DeepMind). The top of the Company → Model Family → Model hierarchy — identity, headquarters country, and slug ID only. Model Release Ledger and Model Families join in on Slug ID; richer country attributes join in via Alpha-2 Code → Country Codes.

Columns
Company String Display name of the AI company/lab (e.g. 'Anthropic', 'OpenAI'). One row per company; this table is the top of the Company → Model Family → Model hierarchy used throughout this dataset.
Headquarters Country String Country where the company is headquartered, as a display name (e.g. 'United States', 'China'). See Alpha-2 Code for the machine-joinable ISO version.
Alpha-2 Code String ISO 3166-1 alpha-2 code for the headquarters country (e.g. 'US', 'CN'). Joins to the Country Codes reference table for richer country-level attributes.
Slug ID String Lowercase, hyphenated unique identifier for the company (e.g. 'anthropic', 'google-deepmind'). This is the foreign key used by Model Release Ledger and Model Families to reference the company — join on this column, not on the company name.

Linked datasets

Datasets that join to AI Companies and Models in the Mycorr graph.