Qwen-3-Swallow-8B-CPT-v0.2: benchmark results
Provider: Alibaba.
Aggregate Rating 1500 ± 34, rank #2361 of 10916 rated models, from 35 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Swallow - Post-trained Japanese - WMT20 En-Ja | 23 | BLEU | 38.8 |
| Swallow - Post-trained Japanese - WMT20 Ja-En | 20.6 | BLEU | 35.8 |
| Swallow - Post-trained English - IFBench | 30.8 | Instruction-Level Strict Accuracy (%) | 34.3 |
| Swallow - Post-trained Japanese - PolyMath High and Top | 20.8 | Accuracy (%) | 34.3 |
| Swallow - Japanese MT-Bench - Roleplay | 60.9 | Judge Score (normalized, %) | 31.3 |
| Swallow - Post-trained English - LiveCodeBench | 34.2 | Pass@1 (%) | 31.3 |
| Swallow - Post-trained Japanese - GPQA | 45.3 | Accuracy (%) | 31.3 |
| Swallow - Post-trained Japanese - JHumanEval | 82.3 | Pass@1 (%) | 31.3 |
| Swallow - Post-trained English - AIME | 40 | Accuracy (%) | 30.6 |
| Swallow - Post-trained English - Average | 54.6 | Average Score (%) | 29.9 |
| Swallow - Post-trained Japanese - MMLU-ProX | 63.6 | Accuracy (%) | 29.9 |
| Swallow - Post-trained English - GPQA Diamond | 49.6 | Accuracy (%) | 28.4 |
Interactive version: theaggregate.ai/model?slug=qwen-3-swallow-8b-cpt-v0-2 · How It Works · Data refreshed daily, snapshot 2026-11-08.