Drift-Bench (State-Oriented, With Clarification) - Premise: leaderboard

Metric: Success rate (%; 200 AgentBench operating-system and database tasks with a premise fault (a false assumption), clarification tools answered by a persona user simulator, mean of 3 trials). Source: arxiv.org. Saturation forecast: Around December 2026. 7 models tracked.

Top models

#ModelScore
1Qwen 3 235B A22B 2507 Instruct68.67
2DeepSeek V3.267.33
3GPT-5.265.84
4GPT-OSS-120B62.67
5Gemini 2.5 Flash58.66
6GLM-4.753.33
7Llama 4 Maverick38.66

Interactive version: theaggregate.ai/benchmark?slug=drift-bench-state-oriented-with-clarification-premise · How It Works · Data refreshed daily, snapshot 2026-09-25.