ResearchBenchmarksTerminal-related CLI agent task on SWE-Bench LiteFollow57.12AccuracyTACO56.267256.488656.7156.9314Apr 21, 2026Evaluation ResultsMethodMethodLinksAccuracyTotal Tokens (M)TACOBackbone=MiniMax-2.5Backbone=MiniMax-2.52026.0457.12270.53BaselineBackbone=MiniMax-2.5Backbone=MiniMax-2.52026.0456.3307.61