Skip to content

Benchmark

SimpleQA Verified

Factuality benchmarks measure how often models get facts right. Scores are percentages. Higher is better.

Models and systems measured
100
Scores recorded
115
Sources
3
IQ weight
1.90

Score distribution

4.7%
Mean
0.451%
Median
14%
Std dev
0.174%
10th percentile
18.9%
90th percentile
0.06%
Min
0.06%73.9%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.90 for chat models, measured over 99 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 Fable 5.2max· Epoch AI Benchmarking Hub data94%
2GPT-6.2 Astramax· Epoch AI Benchmarking Hub data78.5%
3Gemini 4 Carbonhigh· Epoch AI Benchmarking Hub data77.7%
4Gemini 4 Argonhigh· Epoch AI Benchmarking Hub data76%
5GPT-6 Astramax· Epoch AI Benchmarking Hub data75.6%
6Gemini 4 Neonxhigh· Epoch AI Benchmarking Hub data75.2%
7GPT-6.1 Sol· LLM Stats data73.9%
8GPT-6.1 Solmax· Epoch AI Benchmarking Hub data73.9%
9Gemini 3.1 Pro Previewhigh· Epoch AI Benchmarking Hub data73.5%
10Claude Opus 5.5max· Epoch AI Benchmarking Hub data72.2%
11Gemini 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.72.1%
12Gemini 4 Hydrogenxhigh· Epoch AI Benchmarking Hub data70.9%
13Claude Fable 5.1max· Epoch AI Benchmarking Hub data70.8%
14Claude Fable 5xhigh· Epoch AI Benchmarking Hub data70.7%
15Gemini 3.8 Flashhigh· Epoch AI Benchmarking Hub data69.7%
16GPT-5.6 Solmax· Epoch AI Benchmarking Hub data69.7%
17Gemini 3.7 Flashhigh· Epoch AI Benchmarking Hub data69.2%
18Gemini 3 Flashofficial Kaggle leaderboard· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.68.7%
19Gemini 3 Flashhigh· Epoch AI Benchmarking Hub data66.8%
20Gemini 3.5 Flashhigh· Epoch AI Benchmarking Hub data66.2%
21Gemini 3.6 Flashhigh· Epoch AI Benchmarking Hub data66.2%
22GPT 5.5xhigh· Epoch AI Benchmarking Hub data63%
23GPT-6 Solmax· Epoch AI Benchmarking Hub data60.7%
24Muse Spark 1.2xhigh· Epoch AI Benchmarking Hub data60.3%
25Claude Opus 5max· Epoch AI Benchmarking Hub data59.9%
26Muse Spark 1.1· Epoch AI Benchmarking Hub data57.8%
27GPT-6.1 Terramax· Epoch AI Benchmarking Hub data57.3%
28Muse Spark Open 816B A42Bxhigh· Epoch AI Benchmarking Hub data56.6%
29Grok 4.7xhigh· Epoch AI Benchmarking Hub data56%
30Qwen3.7 Max· Epoch AI Benchmarking Hub data55.8%
31Claude Opus 4.8max· Epoch AI Benchmarking Hub data53%
32DeepSeek V4 Pro 0813max· Epoch AI Benchmarking Hub data52.9%
33Qwen3.6 Max Preview· Epoch AI Benchmarking Hub data52%
34Claude Opus 4.7xhigh· Epoch AI Benchmarking Hub data51.7%
35Kimi K3max· Epoch AI Benchmarking Hub data50.6%
36GPT 5high· Epoch AI Benchmarking Hub data50.1%
37OpenAI o3high· Epoch AI Benchmarking Hub data49.4%
38Grok 4.6high· Epoch AI Benchmarking Hub data49.3%
39Grok 4.6xhigh· Epoch AI Benchmarking Hub data48.9%
40Qwen3 Max· Epoch AI Benchmarking Hub data48.7%
41Grok 4.5high· Epoch AI Benchmarking Hub data48.3%
42GPT 5.1high· Epoch AI Benchmarking Hub data48%
43Qwen3.8 Maxxhigh· Epoch AI Benchmarking Hub data47.3%
44Claude Opus 4.6max· Epoch AI Benchmarking Hub data47%
45DeepSeek V4 Promax· Epoch AI Benchmarking Hub data47%
46Claude Sonnet 5.5max· Epoch AI Benchmarking Hub data46.5%
47GPT-5.4 Proxhigh· Epoch AI Benchmarking Hub data46.3%
48Claude Opus 4.532k· Epoch AI Benchmarking Hub data45.7%
49GPT 5.4xhigh· Epoch AI Benchmarking Hub data45.1%
50Doubao Seed 2.1 Prohigh· Epoch AI Benchmarking Hub data44.7%

Showing 1–50 of 114 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
DeepSeek V4.1 Flash BaseEM · 25-shot· Reported by DeepSeek· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.42.3%

Data credits: Epoch AI Benchmarking Hub, CC BY 4.0 (licence) · 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