Posts

Showing posts with the label Alibaba AI

Qwen3.8-Flash-Next: How a 6B Active Model Beats Frontier AI on SWE-bench

Image
Alibaba Releases Qwen3.8-Flash-Next: The Qwen4 Architecture Preview Alibaba Cloud has officially released Qwen3.8-Flash-Next , an experimental open-weight model delivering frontier-level coding performance with unprecedented efficiency. Key Benchmark Highlights: SWE-bench Pro: 62.5% — Beats larger proprietary models in multi-file repository refactoring. LiveCodeBench v6: 91.9% — Exceptional real-time algorithmic reasoning. 195 Tokens/Sec: Sub-210ms Time to First Token (TTFT) via Hybrid Attention (Gated DeltaNet + QSA). 👉 In our in-depth article: Read our detailed analysis on SoftReviewed