Drift-Bench (Service-Oriented, No Clarification) - Premise: leaderboard

Metric: Success rate (%; 150 StableToolBench G1 API tasks with a premise fault (a false assumption), no clarification tools, mean of 3 trials). Source: arxiv.org. Saturation forecast: Around 2030. 7 models tracked.

Top models

#ModelScore
1GLM-4.756.65
2Llama 4 Maverick54.44
3DeepSeek V3.247.11
4Qwen 3 235B A22B 2507 Instruct46.33
5Gemini 2.5 Flash40.44
6GPT-5.236.78
7GPT-OSS-120B34.44

Interactive version: theaggregate.ai/benchmark?slug=drift-bench-service-oriented-no-clarification-premise · How It Works · Data refreshed daily, snapshot 2026-09-25.