Alibaba's Qwen3.8-Flash-Next beats rivals at a fraction of the cost

The Decoder · rss · 2026-08-26

Alibaba releases Qwen3.8-Flash-Next, previewing the Qwen4 architecture. This MoE model activates only 6B of 125B parameters per token. At one-ninth the training cost, it beats larger rivals like DeepSeek-V4-Flash and Claude Opus 4.6 on benchmarks, adding pricing pressure to competitors.

Original post →

More from Models

Models channel →