Agent benchmarks

τ²-bench Core agent benchmark

OpenGPT keeps 8 published rows attached to the exact τ²-bench Core benchmark version and environment shown below.

Tasks: 278
Published rows: 8
Source data date: Aug 6, 2026
Identity reviewed: Current

Evidence scope

v1.0.1 · standard agent · GPT-5.2 low user simulator · 4 trials. base tasks across retail, airline, and telecom. Standard agent, standard domain tools, and a shared GPT-5.2 low-reasoning user simulator. 4 trials per task · 1,112 trajectories per configuration.

How OpenGPT handles this evidence

OpenGPT preserves the publisher's metric, task scope, environment, version, configuration fields, and source date. It does not create a cross-benchmark overall Agent score.

Comparison boundaries

A higher result applies only to the displayed benchmark version and conditions.
Agent results reflect the full system, not the base model alone.
Missing cost, reliability, safety, or latency data remains unreported rather than being inferred.
OpenGPT includes only the eight publisher-submitted rows that share benchmark v1.0.1, the standard scaffold and tools, the same user simulator, the same base task splits, and four trials. Other τ-bench rows remain outside this scope.

v1.0.1 · standard agent · GPT-5.2 low user simulator · 4 trials

Showing 8 of 8 published rows. Every result remains attached to the source benchmark and configuration.