Qwen-3-Swallow-8B-SFT-v0.2: benchmark results
Provider: Alibaba.
Aggregate Rating 1543 ± 34, rank #1799 of 10916 rated models, from 35 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Swallow - Japanese MT-Bench - Math | 97.4 | Judge Score (normalized, %) | 62.7 |
| Swallow - Japanese MT-Bench - Roleplay | 68 | Judge Score (normalized, %) | 62.7 |
| Swallow - Post-trained English - IFBench | 39.5 | Instruction-Level Strict Accuracy (%) | 49.3 |
| Swallow - Japanese MT-Bench - Extraction | 70.9 | Judge Score (normalized, %) | 47.8 |
| Swallow - English MT-Bench - Extraction | 78.2 | Judge Score (normalized, %) | 46.3 |
| Swallow - Japanese MT-Bench - Writing | 61 | Judge Score (normalized, %) | 44.8 |
| Swallow - Japanese MT-Bench - STEM | 62.1 | Judge Score (normalized, %) | 41.8 |
| Swallow - English MT-Bench - Math | 94.8 | Judge Score (normalized, %) | 40.3 |
| Swallow - Japanese MT-Bench - Average | 68.7 | Judge Score (normalized, %) | 40.3 |
| Swallow - Post-trained Japanese - PolyMath High and Top | 25.2 | Accuracy (%) | 38.8 |
| Swallow - Japanese MT-Bench - Reasoning | 66.9 | Judge Score (normalized, %) | 37.3 |
| Swallow - Post-trained Japanese - Average | 44.9 | Average Score (%) | 36.6 |
Interactive version: theaggregate.ai/model?slug=qwen-3-swallow-8b-sft-v0-2 · How It Works · Data refreshed daily, snapshot 2026-11-08.