Swallow - Japanese MT-Bench - Extraction: leaderboard
From the Swallow LLM Leaderboard: Japanese MT-Bench (Nejumi edition, v4 questions with corrected v2 reference answers), two-turn dialogues graded by GPT-5.2 against references; normalized 0-1 upstream, shown as 0-100, higher is better. This board covers the Extraction category: extracting entities, numbers and sentiment from text.
Metric: Judge Score (normalized, %). Source: swallow-llm.github.io. Saturation forecast: Estimated already saturated. 68 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.4 (2026-03-05) (Medium) | 78.7 | #63 (GPT-5.4 (2026-03-05)) |
| 2 | GPT-4o (2024-08-06) | 77.6 | #343 |
| 3 | O3 Mini (Medium) | 77.4 | #282 (O3 Mini) |
| 4 | gemma-4-26B-A4B-it (Thinking) | 77.2 | |
| 5 | GPT-5 Mini (2025-08-07) (Medium) | 77 | #175 (GPT-5 Mini (2025-08-07)) |
| 6 | Qwen3-Swallow-32B-RL-v0.2 (Thinking) | 76.8 | #374 (Qwen3-Swallow-32B-RL-v0.2) |
| 7 | Qwen 3 32B (Thinking) | 76.5 | #449 (Qwen 3 32B) |
| 8 | Gemma 3 27B (IT) | 76.4 | #528 |
| 9 | O3 (2025-04-16) (Medium) | 76.4 | #125 (O3 (2025-04-16)) |
| 10 | GPT-OSS-Swallow-120B-RL-v0.1 (Thinking) | 76.3 | #309 (GPT-OSS-Swallow-120B-RL-v0.1) |
| 11 | GPT-5 (Medium) | 75.8 | #105 (GPT-5) |
| 12 | GPT-4.1 | 75.7 | #258 |
| 13 | Gemma 4 31B (IT) (Thinking) | 75.7 | #178 (Gemma 4 31B (IT)) |
| 14 | llm-jp-4-32B-a3B-Thinking | 75 | #477 |
| 15 | GPT-OSS-20B (Medium) | 74.9 | #523 (GPT-OSS-20B) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=swallow-japanese-mt-bench-extraction · How It Works · Data refreshed daily, snapshot 2026-11-05.