WolfBench: leaderboard

WolfBench mean success rate over Terminal-Bench 2.0’s 89 tasks using Terminus-2. Repeated trials are averaged within each declared agent version, model, timeout and thinking configuration; endpoint and quantization distinctions are retained.

Source: wolfbench.ai.

Interactive version: theaggregate.ai/benchmark?slug=wolfbench · How It Works · Data refreshed daily, snapshot 2026-10-20.