Qwen 3.8 Flash: benchmark results
Alibaba's Qwen3.8-Flash (August 2026), the hosted production version of Qwen3.8-Flash-Next, a 125-billion-parameter MoE activating 6B parameters per token, with a 262K-token native context. Provider: Alibaba. Released 2026-08-26. Access: API.
Aggregate Rating 1824 ± 9, rank #50 of 1656 rated models, from 162 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Grounding Awareness - LangMem - Contextual Selectivity | 0.32 | Contextual Selectivity (0–1) | 100 |
| Grounding Awareness - LangMem - Sensitivity | 0.23 | Sensitivity (0–1) | 100 |
| Grounding Awareness - Mem0 - Sensitivity | 0.13 | Sensitivity (0–1) | 100 |
| Grounding Awareness - Raw History - Sensitivity | 0.14 | Sensitivity (0–1) | 100 |
| Grounding Awareness - Stateless - Sensitivity | 0.1 | Sensitivity (0–1) | 100 |
| Qwen3.8 Omni Report - ClawEval-MM - Pass@3 | 64.4 | Pass@3 (%) | 100 |
| Qwen3.8 Omni Report - ERQA | 72.3 | Score (%) | 100 |
| Qwen3.8 Omni Report - GPQA Diamond | 91.7 | Score (%) | 100 |
| Qwen3.8 Omni Report - RealWorldQA | 88.5 | Score (%) | 100 |
| Qwen3.8 Omni Report - SWE-bench Multilingual | 81 | Score (%) | 100 |
| Qwen3.8 Omni Report - Vision2Web | 64 | Score (%) | 100 |
| RealWorldQA | 88.5 | RealWorldQA (verified) | 99.3 |
Interactive version: theaggregate.ai/model?slug=qwen-3-8-flash · How It Works · Data refreshed daily, snapshot 2026-10-20.