Qwen Unveils 125B Open-Source Model Qwen3.8-Flash-Next

The Qwen3.8-Flash-Next model is an open 125B parameter model. Its architecture is designed to serve as the foundation for the upcoming Qwen4 series.

According to published results, the model competes with closed frontier models in specific tasks, including programming, agent-based workflows, and computer use.

Technical tests showed that a full checkpoint of Qwen/Qwen3.8-Flash-Next can run on a single NVIDIA RTX 4090 GPU. In this configuration, the peak VRAM consumption was 5.95 GB without the use of 4-bit quantization or distillation, utilizing a full checkpoint in bf16.