Research3h ago
Research
Qwen Previews Future Architecture With Flash-Next
The Qwen3.8-Flash-Next model uses a Mixture-of-Experts architecture with 125B total tokens but only 6B active parameters. This release serves as an early preview for the upcoming Qwen4 design. It delivers a significant performance boost over denser models. Practitioners can now test quantized versions via Unsloth to evaluate these new reasoning capabilities.