TAU2-Bench

TAU2-Bench - agentic tool use benchmark for customer service scenarios. Airline domain evaluated with fixes from Claude Opus 4.5 system card.

Category

General Agents

Max Score

—

Score Type

percent

Active

Yes

Model Scores

Scores for TAU2-Bench

Model Score Date Verified
Qwen3.5-4B
78.77%
01.02.2026 Verified
Qwen3.5-122B-A10B
78.27%
01.02.2026 Verified
Qwen3.5-27B
77.65%
01.02.2026 Unverified
Qwen3.5-35B-A3B
80.37%
01.02.2026 Unverified
Qwen3-235B-A22B
52.35%
01.02.2026 Unverified
GPT-5-mini 2025-08-07
66.30%
01.02.2026 Unverified
Gemma4-31B-it
75.06%
02.07.2026 Verified
Gemma4-26B-A4B
64.32%
02.07.2026 Verified
Gemma4-12B
65.31%
02.07.2026 Verified
Gemma4-E4B
32.22%
02.07.2026 Verified
Gemma4-E2B
10.37%
02.07.2026 Verified
Gemma3-27B
0.12%
02.07.2026 Unverified
Spark-X2.5-4B
72.84%
— Verified
Spark-X2.5-1.7B
60.74%
— Verified
Qwen3.5-9B
77.78%
— Verified
Qwen3.5-2B
40.37%
— Verified
LFM2.5-8B-A1B
16.10
— Verified
MiniCPM5-2B
100.00%
— Verified
LFM2.5-2.6B
91.73%
— Verified
granite-4.2-3B
30.62%
— Verified
Nemotron-3-Nano-4B
14.81%
— Verified