Five Qwen3.8-27B models outperform Fable 5 Using GVS5H Harness
133 points by rzk
Five Qwen3.8-27B models have demonstrated comparable performance to Claude Fable 5 on the LiveCodeBench Hard benchmark, achieving similar coding capabilities at a fraction of the cost—approximately one-fifth the price— and can operate on a single GPU. This showcases the efficiency and cost-effectiveness of the Qwen models in high-level coding tasks.