Qwen3.8-27B open weights released, spawns uncensored variants

Qwen3.8-27B, a native multimodal dense model with 27B parameters, outperforms Qwen3.7-Plus and supports 262K native context. Community releases include abliterated and GGUF versions, with one uncensored variant drawing warnings for providing harmful instructions.