Nejumi 4 - jaster (2-shot) - JSICK: leaderboard
JSICK, Japanese translation of the SICK entailment dataset, from the llm-jp-eval jaster suite as run 2-shot by Nejumi Leaderboard 4 (Weights & Biases Japan). Score is exact match (%) on Nejumi's sampled test split; higher is better.
Metric: Exact match (%). Source: nejumi.ai. Saturation forecast: Estimated already saturated. 145 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen3-Swallow-8B-RL-v0.2 (Thinking) | 89 | #602 (Qwen3-Swallow-8B-RL-v0.2) |
| 2 | GPT-5 Nano (2025-08-07) (High) | 88 | |
| 3 | Qwen 3 14B (Reasoning) | 86 | #542 (Qwen 3 14B) |
| 4 | Solar Pro 4 | 86 | #230 |
| 5 | Claude Sonnet 4 (Non-reasoning) | 86 | #207 (Claude Sonnet 4) |
| 6 | O4 Mini (2025-04-16) | 85 | #183 |
| 7 | DeepSeek V3.2 (Thinking) | 85 | #210 (DeepSeek V3.2) |
| 8 | GPT-5.6 Terra (Max) | 85 | #40 (GPT-5.6 Terra) |
| 9 | GPT-4o (2024-11-20) | 84 | #392 |
| 10 | GPT-5.4 Mini (2026-03-17) (High) | 84 | #180 (GPT-5.4 Mini (2026-03-17)) |
| 11 | DeepSeek V4 Flash (0731) (Max) | 84 | #84 (DeepSeek V4 Flash (0731)) |
| 12 | Qwen3-Swallow-32B-RL-v0.2 (Thinking) | 84 | #374 (Qwen3-Swallow-32B-RL-v0.2) |
| 13 | Inkling (Max) | 84 | #144 (Inkling) |
| 14 | GPT-5.6 Sol (Max) | 83 | #22 (GPT-5.6 Sol) |
| 15 | DeepSeek V4 Pro (Max) | 83 | #112 (DeepSeek V4 Pro) |
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-jaster-2-shot-jsick · How It Works · Data refreshed daily, snapshot 2026-11-05.