Skip to content

Benchmark

JMLE2026-Bench (120th JMLE): exam points, text-only questions

Science benchmarks measure how well models answer science questions. Higher scores are better.

Models measured
15
Scores recorded
27
Sources
1
IQ weight
1.98

Score distribution

319
Mean
347
Median
73.8
Std dev
192
10th percentile
376
90th percentile
123
Min
123380
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.98 for chat models, measured over 15 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 4.6· JMLE2026-Bench and IgakuQA119 data docs380
2Claude Sonnet 4.6· JMLE2026-Bench and IgakuQA119 data docs378
3Gemini 3.1 Pro Preview· JMLE2026-Bench and IgakuQA119 data docs378
4GPT 5.2· JMLE2026-Bench and IgakuQA119 data docs376
5Qwen3.5 35B A3B· JMLE2026-Bench and IgakuQA119 data docs370
6Qwen3.5 397B A17B· JMLE2026-Bench and IgakuQA119 data docs370
7Qwen3.5 122B-A10B· JMLE2026-Bench and IgakuQA119 data docs367
8Qwen3.5 27B· JMLE2026-Bench and IgakuQA119 data docs365
9GPT OSS 120Bhigh· JMLE2026-Bench and IgakuQA119 data docs362
10Qwen3.5 27Bno-think· JMLE2026-Bench and IgakuQA119 data docs359
11Qwen3.5 122B-A10Bno-think· JMLE2026-Bench and IgakuQA119 data docs355
12Qwen3.5 35B A3Bno-think· JMLE2026-Bench and IgakuQA119 data docs354
13Qwen3 32B· JMLE2026-Bench and IgakuQA119 data docs349
14GPT OSS 120Bmedium· JMLE2026-Bench and IgakuQA119 data docs347
15Qwen3.5 9B· JMLE2026-Bench and IgakuQA119 data docs340
16GPT OSS 120Blow· JMLE2026-Bench and IgakuQA119 data docs333
17GPT OSS 20bhigh· JMLE2026-Bench and IgakuQA119 data docs332
18GPT OSS 20bmedium· JMLE2026-Bench and IgakuQA119 data docs330
19Qwen3.5 9Bno-think· JMLE2026-Bench and IgakuQA119 data docs325
20Qwen3.5 4B· JMLE2026-Bench and IgakuQA119 data docs323
21Qwen3 32Bno-think· JMLE2026-Bench and IgakuQA119 data docs321
22GPT OSS 20blow· JMLE2026-Bench and IgakuQA119 data docs299
23Qwen3.5 4Bno-think· JMLE2026-Bench and IgakuQA119 data docs290
24Qwen3.5 2Bno-think· JMLE2026-Bench and IgakuQA119 data docs192
25Qwen3.5 2B· JMLE2026-Bench and IgakuQA119 data docs161
26Qwen3.5 0.8Bno-think· JMLE2026-Bench and IgakuQA119 data docs134
27Qwen3.5 0.8B· JMLE2026-Bench and IgakuQA119 data docs123

Data credits: JMLE2026-Bench and IgakuQA119 by Naoto Iwase; results licence not stated (code MIT, question data CC BY 4.0).