Skip to content

Benchmark

MMMLU

Multilingual benchmarks compare model performance across languages. Scores are percentages. Higher is better.

Models and systems measured
49
Scores recorded
163
Sources
2
IQ weight
2.48

Score distribution

79.8%
Mean
81.3%
Median
10.1%
Std dev
68.3%
10th percentile
89.6%
90th percentile
38.4%
Min
38.4%92.6%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.48 for chat models, measured over 45 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
1Gemini 3.1 Pro Previewsetting not named by the source· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.92.6%
2Gemini 3.1 Pro Preview· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim92.6%
3Gemini 3 Flash· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim91.8%
4Gemini 3 Flashmultilingual Q&A· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.91.8%
5Gemini 3 Pro Preview· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim91.8%
6Gemini 3 Pro PreviewGemini 3 Pro Model Card · results as of November 2025· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.91.8%
7Gemini 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.91.8%
8Claude Opus 4.7· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim91.5%
9Claude Opus 4.714 non-English languages · adaptive thinking · max effort · avg@3· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.91.5%
10Claude Opus 4.6· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim91.1%
11Claude 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.91.1%
12Claude Opus 4.5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim90.8%
13Claude 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.90.8%
14Claude Opus 4.514 non-English languages · 64K thinking · 200K context · avg@10· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.90.8%
15Qwen3.7 Max· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim90.3%
16Qwen3.7 MaxQwen3.7-Max release · recommended reasoning-effort=xhigh system prompt· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.90.3%
17GPT 5.2· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim89.6%
18GPT 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.89.6%
19GPT 5.2xhigh· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.89.6%
20Claude Opus 4.1· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim89.5%
21Claude 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.89.5%
22Qwen3.6 Plus· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim89.5%
23Qwen3.6 PlusQwen official release table · thinking mode unless benchmark requires otherwise· Reported by Alibaba Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.89.5%
24Claude Sonnet 4.6· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim89.3%
25Claude Sonnet 4.614 non-English 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.89.3%
26Claude Sonnet 4.5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim89.1%
27Claude Sonnet 4.5launch-page setting· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.89.1%
28Qwen3.7 Plus· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim89%
29Qwen3.7 PlusQwen3.7-Plus release · recommended reasoning-effort=xhigh system prompt· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.89%
30Gemini 3.1 Flash Lite· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim88.9%
31Gemini 3.1 Flash LiteGemini 3.1 Flash-Lite Model Card · high thinking · default sampling · single attempt/pass@1· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.88.9%
32Claude Opus 4· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim88.8%
33Claude Opus 414 non-English languages · extended thinking up to 64K tokens· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.88.8%
34Qwen3.5 397B A17B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim88.5%
35Qwen3.5 397B A17BQwen3.5 release/model card · published setting· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.88.5%
36Gemma 4 31B IT· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim88.4%
37Gemma 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.88.4%
38o1· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim87.7%
39Claude Opus 414 non-English languages · no extended thinking· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.87.4%
40GPT 4.1· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim87.3%
41Qwen 3 235B A22B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim86.7%
42Qwen3.5 122B-A10B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim86.7%
43Claude Sonnet 4· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim86.5%
44Claude Sonnet 414 non-English languages · extended thinking up to 64K tokens· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.86.5%
45Gemma 4 26B A4B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim86.3%
46Gemma 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.86.3%
47Claude Sonnet 3.7· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim86.1%
48GPT OSS 120Bofficial model card Table 2 · Spanish · reasoning high · translated-answer parser· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.85.9%
49Qwen3.5 27B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim85.9%
50Mistral Largelatest· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim85.5%

Showing 1–50 of 159 results.

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
Qwen3.5-397B-A17B-BaseQwen3.5 release/model card · base-model table· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.85.8%
Qwen3.7-Plus-BaseQwen3.8-Flash-Next release/model card · base-model table· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.84.5%
Qwen3.8-27B-BaseQwen3.8-Flash-Next release/model card · base-model table· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.79.7%
Qwen3.8-Flash-Next-BaseQwen3.8-Flash-Next release/model card · base-model table· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.84.9%

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. · Data by LLM Stats