Predictability on τ-bench (airline, original)

How well does the agent's expressed confidence predict whether it will answer correctly? See full definition →

• Consistency
• Predictability
• Robustness
• Safety

Calibration Curves

Accuracy-Coverage Curves

Agent Leaderboard — Predictability

#	Agent	Acc	Predictability Agg	Cal	AUROC	Brier	Overall
1	Claude Opus 4.5	58.4%	0.69	0.71	0.67	0.69	0.80
2	Claude Sonnet 4.5	54.4%	0.66	0.68	0.61	0.66	0.79
3	GPT-5.2 (xhigh)	51.6%	0.65	0.68	0.62	0.65	0.76
4	O1	49.6%	0.64	0.74	0.50	0.64	0.76
5	Gemini 3.0 Pro	58.8%	0.60	0.60	0.52	0.60	0.77
6	Gemini 2.5 Pro	52.8%	0.56	0.56	0.58	0.56	0.74
7	GPT-5.2	42.0%	0.55	0.55	0.56	0.55	0.75
8	Claude 3.7 Sonnet	43.6%	0.54	0.54	0.53	0.54	0.72
9	Gemini 2.5 Flash	47.2%	0.52	0.52	0.54	0.52	0.70
10	Claude 3.5 Haiku	29.6%	0.46	0.45	0.44	0.46	0.68
11	GPT-4 Turbo	35.6%	0.38	0.38	0.47	0.38	0.69
12	Gemini 2.0 Flash	32.0%	0.38	0.36	0.61	0.38	0.67
13	GPT-4o Mini	21.3%	0.32	0.29	0.48	0.32	0.67