Skip to content

Benchmark

Global MMLU (GMMLU)

Scores are percentages. Higher is better.

Models and systems measured
9
Scores recorded
27
Sources
1
IQ weight
0.91

Score distribution

86.9%
Mean
88.1%
Median
4.9%
Std dev
80.3%
10th percentile
92.5%
90th percentile
76.6%
Min
76.6%94.3%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.91 for chat models, measured over 8 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#Model or systemScore
1Claude Opus 5.5Opus 5.5 System Card · adaptive thinking · max effort unless stated· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.94.3%
2Claude Fable 5.1Opus 5.5 System Card · adaptive thinking · max effort unless stated· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.94%
3Claude Sonnet 4.6English · adaptive thinking · max effort· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.92.9%
4Claude Opus 5Opus 5.5 System Card · adaptive thinking · max effort unless stated· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.92.5%
5Claude Sonnet 5.5adaptive thinking · max effort · default sampling · 5-trial mean · context ≤1M unless noted · 42-language average · adaptive thinking · max · single trial· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.92.1%
6Claude Sonnet 4.6high-resource average · adaptive thinking · max effort· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.91%
7Claude Sonnet 4.6mid-resource average · adaptive thinking · max effort· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.90.2%
8Claude Opus 4.7average across all evaluated languages· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.89.9%
9Claude Sonnet 5Opus 5.5 System Card · adaptive thinking · max effort unless stated· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.89.2%
10Claude Sonnet 4.6Nepali · adaptive thinking · max effort· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.89.1%
11Claude Sonnet 5overall average · adaptive thinking · max effort · 1 trial· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.89%
12Claude Sonnet 4.6overall average · all languages · adaptive thinking · max effort· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.88.7%
13Claude Sonnet 4.6overall average · Sonnet 5 card rerun· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.88.4%
14Claude Sonnet 4.6Sinhala · adaptive thinking · max effort· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.88.1%
15Claude Haiku 5.5adaptive thinking · max effort · default sampling · 5-trial mean · context ≤1M unless noted · 42-language average · adaptive thinking · max · single trial· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.87.8%
16Claude Sonnet 4.6Swahili · adaptive thinking · max effort· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.87%
17Claude Sonnet 4.6Kyrgyz · adaptive thinking · max effort· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.86.9%
18Claude Sonnet 4.6Amharic · adaptive thinking · max effort· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.86.7%
19Claude Sonnet 4.6Hausa · adaptive thinking · max effort· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.84.1%
20Claude Sonnet 4.6Malagasy · adaptive thinking · max effort· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.83.9%
21Claude Sonnet 4.6low-resource average · adaptive thinking · max effort· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.83.8%
22Claude Sonnet 4.6Somali · adaptive thinking · max effort· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.82.3%
23Claude Sonnet 4.6Shona · adaptive thinking · max effort· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.82.2%
24Claude Sonnet 4.6Yoruba · adaptive thinking · max effort· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.80.3%
25Claude Sonnet 4.6Chichewa · adaptive thinking · max effort· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.78.8%
26Claude Sonnet 4.6Igbo · adaptive thinking · max effort· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.76.7%

Not ranked

The source marks these results as descriptive, not capability scores, so we list them by name instead of ranking them.

Model or systemScore
MiMo V2 Flash Baseofficial Base table · 5-shot · GlobalMMLU· Reported by Xiaomi MiMo· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.76.6%

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.