Qwen3.8-Flash-Next Official Release ReportProvider report

Name: Qwen3.8-Flash-Next Official Release Report
Published: 2026-08-26

2026-08-26Based on the two complete performance comparison tables in Qwen's official August 26, 2026 release article (five text columns and four multimodal columns). This snapshot includes all 12 source rows from the text table and all 10 source rows from the multimodal table. It splits the compound cells for Agents' Last Exam, ClawEval-MM, OSWorld 2.0, MathVision, and CharXiv (RQ), producing 27 metrics and 109 available scores. Comparison-model scores are preserved as reported, and '--' cells do not create scores. CoWorkBench and RecreationBench are Qwen internal benchmarks. The base-model comparison table (Flash-Next-Base / 27B-Base / 3.7-Plus-Base) covers base variants that are not catalogued, so it is excluded.

Back to data sources

Qwen3.8-Flash-Next Official Release ReportProvider report

View source

2026-08-26Based on the two complete performance comparison tables in Qwen's official August 26, 2026 release article (five text columns and four multimodal columns). This snapshot includes all 12 source rows from the text table and all 10 source rows from the multimodal table. It splits the compound cells for Agents' Last Exam, ClawEval-MM, OSWorld 2.0, MathVision, and CharXiv (RQ), producing 27 metrics and 109 available scores. Comparison-model scores are preserved as reported, and '--' cells do not create scores. CoWorkBench and RecreationBench are Qwen internal benchmarks. The base-model comparison table (Flash-Next-Base / 27B-Base / 3.7-Plus-Base) covers base variants that are not catalogued, so it is excluded.