spb/wp13_uqo

1\begin{table}[htbp]2\centering3\caption{Model layer at the audited commit. The reference path targets the Anthropic Messages API;4adaptive extended thinking is enabled only for the models listed as supporting it. OpenAI and5Google models are served through an OpenAI-compatible adapter that receives the first 128 tools of6the 245-tool catalogue. Model identifiers are reproduced verbatim from the client configuration7(\texttt{client/src/lib/models.ts}); the Anthropic identifiers were checked against the provider's8current model list, the others were not audited.}9\label{tab:models}10\begin{threeparttable}11\footnotesize12\begin{tabular}{@{}llccl@{}}13\toprule14Provider path & Model identifier & Adaptive thinking & Context & Role in selector \\15\midrule16Anthropic (reference) & \texttt{claude-fable-5}            & yes & 1M   & default \\17Anthropic             & \texttt{claude-opus-4-8}           & yes & 1M   & long-horizon agentic \\18Anthropic             & \texttt{claude-opus-4-7}           & yes & 1M   & balanced \\19Anthropic             & \texttt{claude-opus-4-6}           & yes & 1M   & fast and capable \\20Anthropic             & \texttt{claude-sonnet-4-6}         & yes & 1M   & general purpose \\21Anthropic             & \texttt{claude-haiku-4-5-20251001} & no  & 200K & lowest latency \\22OpenAI-compatible     & twelve OpenAI identifiers          & n/a & --   & not audited \\23OpenAI-compatible     & four Google identifiers            & n/a & --   & not audited \\24\bottomrule25\end{tabular}26\begin{tablenotes}[flushleft]27\footnotesize28\item \textit{Notes.} Per-response generation is capped at 64{,}000 tokens and the temperature is29set to~1 on the reference path; the summarisation call of Algorithm~\ref{alg:context} uses a cap30of 2{,}000 tokens and a temperature of 0.5. The Beamer restructuring call31(Section~\ref{subsec:docs}) is pinned to \texttt{claude-opus-4-8} irrespective of the selected32model. Cost is computed from a single fixed schedule (\$3 per million input tokens, \$15 per million33output tokens) for every model, so the ledger does not reflect per-model prices34(Section~\ref{subsec:accounting}).35\end{tablenotes}36\end{threeparttable}37\end{table}38