Model groups

All models all_models

Every model eligible for the capability. Applies no allowlist — the pool is computed per call, so there is no membership list to go stale. This is the default pool a task runs in when no group is named, and it exists so that pool has a name to be priced, latched and reported against.

This group is every active model.

Tasks on this group: Factual Claim Refinement, Profiled Document Structure Extraction, Atomic Fact Claim Extraction, Research Query Generation, Multi Perspective Decision Synthesis, Metadata Paragraph Rewriting, Topic Sequence Optimization, Bench Hypothesis Post Generation, Capability Category Matching, Authoritative Source Selection, Promotional Campaign Generation, Engagement Opportunity Triage, Public Response Generation, Topic Cluster Labeling, Social Post Portfolio Selection, Topic Section Assignment, Catalyst And Scenario Analysis, Report Outline Generation, Content Set Relevance Scoring, Research Region Identification, Profile Pool Matching, Content To Image Prompt Generation, Model Specific Prompt Adaptation, Voice Profile Generation, Evidence Grounded Claim Generation, Visual Theme Configuration Generation, Thematic Topic Discovery, Structured Content Summarization, Profiled Document Section Analysis, Claim Continuity Classification, Newsletter Copy Generation, Persona Policy Eligibility Check, Publication Title Package Generation, Community Policy Evaluation, Report Executive Summary Generation, Short Post Batch Relevance Scoring, Profiled Document Analysis, Language Identification, Public Engagement Reply Review, Generated Content Relevance Scoring, Markdown Newline Repair, Publication Section Taxonomy Generation, Community Content Promotion Generation, Research Query Validation, Subject Configuration Generation, Promotional Message Generation, Batch Text Translation, Topic Taxonomy Matching, Bench Article Hypothesis Generation, Reference Preserving Analytical Writing, Section Prompt Generation, Research Community Selection, Content Domain Suggestion, Structured Output Extraction.

US models us_models

Every active model from a US company. No cost filter — the flagship tier is included.

US workhorse us_workhorse

The US pool minus the flagship tier: a strict subset of us_models, and the US-only counterpart of workhorse.

Workhorse workhorse

Every active model except the flagship tier. The default working pool for cost-sensitive routing; premium models are the exception, named explicitly when wanted.