Skip to content

Benchmark

τ²-bench Telecom

Scores are percentages. Higher is better.

Models measured
21
Scores recorded
27
Sources
1
IQ weight
2.31

Score distribution

89%
Mean
97.8%
Median
15.4%
Std dev
64.4%
10th percentile
98.7%
90th percentile
43%
Min
43%99.3%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.31 for chat models, measured over 21 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1Claude Opus 4.6Opus 4.6 System Card capability summary · adaptive thinking · max effort · default sampling · 5 trials unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.99.3%
2Gemini 3.1 Pro Preview· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.99.3%
3GPT 5.4reasoning enabled· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.98.9%
4GPT 5.2Opus 4.6 System Card capability summary · adaptive thinking · max effort · default sampling · 5 trials unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.98.7%
5GPT 5.2reasoning enabled· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.98.7%
6GPT 5.2xhigh· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.98.7%
7Claude Opus 4.5Claude Opus 4.1 user simulator · prompt addendum· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.98.2%
8Claude Opus 4.5Opus 4.6 System Card capability summary · adaptive thinking · max effort · default sampling · 5 trials unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.98.2%
9Claude Sonnet 4.5Claude Opus 4.1 user simulator · prompt addendum· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.98%
10Claude Sonnet 4.5Opus 4.6 System Card capability summary · adaptive thinking · max effort · default sampling · 5 trials unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.98%
11Gemini 3 Pro PreviewOpus 4.6 System Card capability summary · adaptive thinking · max effort · default sampling · 5 trials unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.98%
12GPT 5.5original prompts · no prompt tuning · xhigh· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.98%
13Claude Sonnet 4.6adaptive thinking · max effort · 10 trials· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.97.9%
14MiniMax M2.5search/tool figure· Reported by MiniMax· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.97.8%
15MiMo V2 ProXiaomi MiMo-V2-Pro release table · Telecom· Reported by Xiaomi MiMo· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.96.8%
16DeepSeek 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.96.2%
17MiMo V2 FlashDeepSeek-V3.2 user agent · Telecom· Reported by Xiaomi MiMo· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.95.3%
18Trinity 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.94.7%
19GPT-5.4 minixhigh · telecom domain· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.93.4%
20GPT 5.4 Nanoxhigh · telecom domain· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.92.5%
21Claude Haiku 4.5launch-page setting· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.83%
22Nova 2 LiteAmazon Nova 2 Lite official launch chart / technical report · Artificial Analysis setup, 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.76%
23Gemma 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.69.3%
24Nemotron 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.64.4%
25GPT 5.4no reasoning· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.64.3%
26GPT 5.2no reasoning· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.57.2%
27Gemma 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.43%

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.