Open Japanese LLM v2 - JamC-QA: leaderboard
From LLM-jp's Open Japanese LLM Leaderboard v2, which runs open-weight models through llm-jp-eval v2 with vLLM. JamC-QA (SB Intuitions) is four-choice question answering on Japan-specific knowledge, answered with a single digit and scored by exact match. Runs that returned no parseable answer are omitted. Upstream 0-1 scores are shown as 0-100, higher is better.
Metric: Exact Match (%). Source: huggingface.co. Saturation forecast: Around 2029. 25 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Llama-3.3-Swallow-70B-Instruct-v0.4 | 58.77 | #540 |
| 2 | llm-jp-4.1-33B (Thinking) | 55.61 | #256 (llm-jp-4.1-33B) |
| 3 | Qwen 3.5 27B (Non-reasoning) | 54.83 | #192 (Qwen 3.5 27B) |
| 4 | Qwen 3.6 27B (Non-reasoning) | 54.66 | #182 (Qwen 3.6 27B) |
| 5 | llm-jp-4-32B-a3B-Thinking | 53.66 | #477 |
| 6 | llm-jp-4.1-32B-a3B (Thinking) | 52.75 | #366 (llm-jp-4.1-32B-a3B) |
| 7 | llm-jp-3.1-13B-instruct4 | 49.76 | #843 |
| 8 | GPT-OSS-120B | 49.37 | #346 |
| 9 | llm-jp-4-8B-Thinking | 49.33 | #529 |
| 10 | llm-jp-4.1-8B (Thinking) | 48.25 | #408 (llm-jp-4.1-8B) |
| 11 | gemma-4-26B-A4B-it (Thinking) | 48.03 | |
| 12 | Gemma 3 27B (IT) | 47.94 | #528 |
| 13 | Qwen 3 32B (Non-reasoning) | 45.3 | #449 (Qwen 3 32B) |
| 14 | Llama 3.3 70B Instruct | 44.87 | #541 |
| 15 | Qwen 3 8B (Thinking) | 32.01 | #678 (Qwen 3 8B) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=open-japanese-llm-v2-jamc-qa · How It Works · Data refreshed daily, snapshot 2026-11-05.