LLM-jp4 Original Release Judge - MT-BenchJA: leaderboard
Officialllm-jp4card: gpt-5.4-2026-03-05 judge;3inference/evaluationround average. Exact API snapshots andlow/medium effort retained. JA/ENMTBench80questions;llm-jp-instructions400testquestions. AnswerCarefullypolicybehavior excluded fromcapability intake. Metric: Judge score. Higher is better. Source: https://huggingface.co/llm-jp/llm-jp-4-33b-thinking/raw/main/README.md
Metric: Judge score. Source: huggingface.co. Saturation forecast: Around January 2027. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.4 (2026-03-05) (High) | 8.98 | #57 (GPT-5.4 (2026-03-05)) |
| 2 | GPT-5.4 (2026-03-05) (Medium) | 8.87 | #57 (GPT-5.4 (2026-03-05)) |
| 3 | GPT-5.4 (2026-03-05) (Low) | 8.87 | #57 (GPT-5.4 (2026-03-05)) |
| 4 | GPT-OSS-20B (Medium) | 7.33 | #516 (GPT-OSS-20B) |
| 5 | GPT-4o (2024-08-06) | 7.29 | #331 |
| 6 | GPT-OSS-20B (Low) | 7.21 | #516 (GPT-OSS-20B) |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #7 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=llm-jp4-original-release-judge-mt-benchja · How It Works · Data refreshed daily, snapshot 2026-10-17.