Skip to content

Benchmark

τ²-bench Airline

Scores are percentages. Higher is better.

Models measured
10
Scores recorded
13
Sources
1
IQ weight
0.60

Score distribution

71.3%
Mean
70%
Median
9%
Std dev
63.6%
10th percentile
87.8%
90th percentile
56.3%
Min
56.3%88%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.60 for chat models, measured over 10 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1Trinity Large ThinkingTrinity-Large-Thinking Model Card· Reported by Arcee AI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.88%
2Claude Opus 4.5Anthropic-corrected task setup and grading · fixes submitted to benchmark authors· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.87.8%
3Claude Sonnet 4.5Anthropic-corrected task setup and grading · fixes submitted to benchmark authors· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.77.4%
4Gemma 4 26B A4BGemma 4 Technical Report · Table 5 · thinking mode unless noted· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.76%
5Gemma 4 31B ITGemma 4 Technical Report · Table 5 · thinking mode unless noted· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.75%
6Claude Opus 4.5original task set · Table 2.8.A· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.70.1%
7Claude Sonnet 4.5original task set· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.70%
8Claude Opus 4.5original task set · Table 2.3 footnote· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.67.9%
9MiMo V2 FlashDeepSeek-V3.2 user agent · Airline· Reported by Xiaomi MiMo· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.66%
10Nova 2 LiteAmazon Nova 2 Lite official launch chart / technical report · Verified · avg@3, v1· Reported by Amazon Web Services, Amazon AGI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.64.8%
11DeepSeek V3.2Thinking mode · temperature 1.0 · 128K context · model itself as user agent· Reported by DeepSeek· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.63.8%
12Claude Haiku 4.5original task set · launch-page setting· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.63.6%
13Nemotron 3 Super 120B A12BNeMo Evaluator / mostly NeMo Skills · default prompt· Reported by NVIDIA· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.56.3%

Data credits: BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links.