Swallow - Pre-trained Japanese - JHumanEval Leaderboard · The Aggregate

Swallow - Pre-trained Japanese - JHumanEval: leaderboard

Metric: Pass@1 (%). Source: swallow-llm.github.io. 50 models tracked.

Top models

#	Model	Score
1	Qwen3-14B-Base	70.9
2	Qwen 2.5 72B	64.8
3	Qwen3-30B-A3B-Base	64.1
4	Qwen 2.5 32B	63.7
5	Qwen3-8B-Base	59.5
6	Qwen 2.5 14B	55
7	Qwen3-4B-Base	53.7
8	Qwen 2.5 7B	50.7
9	Gemma 2 27B	49
10	gemma-3-27B-pt	47.3
11	Llama 3.1 70B	46.2
12	Falcon3-10B-Base	42.6
13	Qwen 2.5 3B	40.4
14	gemma-3-12B-pt	38.5
15	Falcon3-7B-Base	36.1

Interactive version: theaggregate.ai/benchmark?slug=swallow-pre-trained-japanese-jhumaneval · How It Works · Data refreshed daily, snapshot 2026-09-19.