Models in the corpus
15 models, three tiers,
one apples-to-apples corpus.
Hundreds of contributor teams running every major frontier model in production. Tier reflects benchmark recency; contributor multiplier is a function of model tier and domain mix.
ModelTierCoverage
Claude Opus 4.8Anthropic
Frontier
100K+ traces
Claude Opus 4.7Anthropic
Frontier
11M+ traces
GPT‑5.5OpenAI
Frontier
10M+ traces
Gemini 4 UltraGoogle
Frontier
5M+ traces
Grok 5xAI
Frontier
9M+ traces
Claude Sonnet 4.5Anthropic
Stable
5M+ traces
GPT‑5OpenAI
Stable
1M+ traces
Gemini 4 ProGoogle
Stable
2M+ traces
DeepSeek V4DeepSeek
Stable
3M+ traces
Llama 5 (405B)Meta
Stable
3M+ traces
Claude Opus 4.6Anthropic
Legacy
2M+ traces
Claude Sonnet 4.6Anthropic
Legacy
3M+ traces
GPT‑5 miniOpenAI
Legacy
2M+ traces
Gemini 3 ProGoogle
Legacy
1M+ traces
Claude Haiku 4Anthropic
Legacy
1M+ traces
Coverage updates daily as contributors approve new batches. Models that lose production traffic age out of the corpus and into the archive.