Alibaba open-sources Qwen3.8 with 2.4T parameters and local-friendly 27B variant

Alibaba released Qwen3.8 as open-weight, featuring a massive 2.4T-parameter model and a 27B variant that runs on consumer GPUs. This challenges the narrative that only US labs lead open-weight releases and provides enterprises with a powerful, deployable alternative. The 27B model requires ~54GB BF16, 27GB FP8, or 14-16GB 4-bit, making it viable on 24GB consumer cards. This is a significant foundation model release in the open-source ecosystem.