Alibaba Cloud released Qwen3.8-27B, a compact 27B-parameter LLM demonstrating significant performance improvements over its predecessor, with coding capabilities comparable to Claude Opus 4.6 (scoring 73.0 on Terminal Bench, 61.7 on SWE-bench Pro). The model is now available for download and local deployment on consumer GPUs; API access is planned, though third-party providers like AkashML are already offering service at $0.45/$3.20 per million tokens—relatively expensive for the model's size, making self-hosting more attractive for SMEs.