Skip to content
Bencher

TestsBenchmarks · Tool use

τ³-bench Banking

τ³-bench banking domain with agentic retrieval (Sierra).

% solvedHigher is betterThe test's websiteOfficial page ↗

The best 3 · 3 models tested

Top 3 · best setting per model · 3 models, 3 results (0 independent)

Measured by independent testersIndependent Reported by the makerVendor-reported
τ³-bench Banking leaderboard
Thinking levelSetting
1Muse Glimmer 30BMeta (Meta Superintelligence Labs, Muse)23.5%HighHigh reasoningMaker's own figureVendor-reportedMeta ↗—10 Aug 2026
2Mistral Medium 3.5Mistral AI13.4%Highmaximum reasoning settings (reasoning_effort=high)Maker's own figureVendor-reportedMistral AI ↗—22 May 2026
3Mistral Small 4Mistral AI7.0%Defaultreasoning setting not statedMaker's own figureVendor-reportedMistral AI ↗—22 May 2026