What does JSQuAD measure?

JSQuAD is a knowledge benchmark in the BenchGecko catalog. 12 AI models have been tested on it. Scores range from 13.9 to 89.9 out of 100.

Which model leads on JSQuAD?

Qwen2 VL 7B Instruct from Alibaba leads JSQuAD with a score of 89.9. The median score across 12 tested models is 86.3.

No · the top score is 89.9 out of 100 (90%). There is still meaningful room for improvement on JSQuAD.

Yes · JSQuAD scores correlate 0.87 with LLM-JP · Overall across 12 shared models. Models that do well on JSQuAD tend to do well on LLM-JP · Overall.

BenchGecko pulls updates daily. New model scores on JSQuAD appear as soon as they are published by Epoch AI or the model provider.

Benchmark · KnowledgeSettled

Name: JSQuAD Benchmark
Creator: BenchGecko
License: https://creativecommons.org/licenses/by/4.0/

Updated 2025-01-20

Models tested

Top score

89.9

Qwen2 VL 7B Instruct

Median

86.3

min 13.9

Top-5 spread

σ 0.2

Settled

Best score over time · one chart, every benchmark

Chart type

Only 6 models have been tested on JSQuAD · not enough history to compute a frontier yet.

Pink dots = frontier records · 1 totalClick to open model page

12 models tested · sorted by score

#	Model	Score	Price
1	Qwen2 VL 7B Instruct· Alibaba	89.9	—
2	DeepSeek R1 Distill Qwen 14B· DeepSeek	89.8	—
3	Qwen2 7B Instruct· Alibaba	89.6	—
4	Meta Llama 3 8B Instruct· Meta	89.5	—
5	DeepSeek R1 Distill Qwen 32B· DeepSeek	89.2	—
6	Meta Llama 3 8B· Meta	88.9	—
7	Gemma 2 2b It· Google DeepMind	83.8	—
8	Llama 2 7b Chat Hf· Meta	83.0	—
9	DeepSeek R1 Distill Llama 8B· DeepSeek	80.2	—
10	Llama 2 7b Hf· Meta	79.9	—
11	DeepSeek R1 Distill Qwen 7B· DeepSeek	74.2	—
12	HF SmolLM2 135M Instruct· Hugging Face TB	13.9	—