Open Japanese LLM - ALT J TO E Comet Wmt22: leaderboard
The "ALT J TO E Comet Wmt22" subtask of Open Japanese LLM. A task from the Open Japanese LLM Leaderboard, evaluating models on Japanese-language knowledge, reasoning and understanding.
Metric: Score (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 418 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Llama-3.3-Swallow-70B-Instruct-v0.4 | 89.23 | #529 |
| 2 | aya-expanse-32B | 89.13 | #692 |
| 3 | Llama 3.1 70B | 89.02 | #592 |
| 4 | Llama-3.1-Tulu-3-70B | 88.93 | #482 |
| 5 | Llama 3 70B | 88.88 | #631 |
| 6 | Llama 3.1 Nemotron 70B Instruct | 88.82 | #554 |
| 7 | Llama 3.1 70B Instruct | 88.81 | #561 |
| 8 | Llama 3.3 70B Instruct | 88.8 | #534 |
| 9 | Qwen 2.5 32B | 88.79 | #546 |
| 10 | cogito-v1-preview-qwen-14B | 88.73 | |
| 11 | Qwentile2.5-32B-Instruct | 88.65 | #339 |
| 12 | aya-23-35B | 88.62 | #883 |
| 13 | Qwen 2.5 14B | 88.62 | #725 |
| 14 | QwQ 32B-Preview | 88.61 | #651 |
| 15 | Qwen2.5-32B-Instruct-CFT | 88.61 | #375 |
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-alt-j-to-e-comet-wmt22 · How It Works · Data refreshed daily, snapshot 2026-10-20.