Nejumi 4 - GLP - Information Retrieval: leaderboard
The JSQuAD reading comprehension set as packaged in the jaster harness, a Japanese version of SQuAD in which the model answers a question using a passage supplied in the prompt. Answers are graded by character-level F1 against the reference span and averaged over the 0-shot and 2-shot runs, reported in percent where higher is better. Source: github.com/wandb/llm-leaderboard.
Metric: Score (%). Source: nejumi.ai. Saturation forecast: Around January 2028. 139 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.6 27B | 82.46 | #170 |
| 2 | Gemini 3.6 Flash | 82.34 | #51 |
| 3 | Qwen 3 Next 80B A3B Instruct | 82.31 | #368 |
| 4 | Hy4 preview | 82.3 | #38 |
| 5 | Claude Opus 5 | 82.24 | #12 |
| 6 | Claude Fable 5 | 82 | #16 |
| 7 | Gemini 3 Pro (Preview) | 81.77 | #66 |
| 8 | Gemini 3.5 Flash | 81.68 | #60 |
| 9 | ELYZA-Shortcut-1.0-Qwen-32B | 81.61 | #547 |
| 10 | Inkling | 81.51 | #126 |
| 11 | Gemma 4 31B (IT) | 81.49 | #176 |
| 12 | MiMo-V2.5-Pro | 81.4 | #158 |
| 13 | GLM-5.3 Flash | 81.39 | #65 |
| 14 | Gemini 3.1 Pro (Preview) | 81.14 | #58 |
| 15 | GPT-5.4 Mini (2026-03-17) | 81.12 | #169 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=nejumi-4-glp-information-retrieval · How It Works · Data refreshed daily, snapshot 2026-10-20.