The user tested Unsloth's Qwen 3.8 Flash Next (Q4_K_XL) model against Anthropic's Opus 5.5 xhigh on a Flutter‑to‑SwiftUI migration task, using a high‑effort configuration on a 128 GB RAM Strix Halo laptop with Gufo on Windows and OpenCode v2 on macOS. Both models received the same prompt to rewrite several app screens from scratch, and the Unsloth model demonstrated performance comparable to Opus 5.5 xhigh. The experiment highlights the strong capabilities of the unsloth‑quantized Qwen model for complex code‑translation workloads.