DeepSeek, the Chinese AI lab, recently refreshed V4-Flash, moving it out of preview, after documentation promised the model topped the V4-Pro preview on coding and agentic benchmarks. V4-Flash runs $0.14 per million input tokens and $0.28 per million output, while V4-Pro runs $0.435 and $0.87, making Flash almost exactly one-third the price on both input and output. The author tested both models on three professional coding jobs using the Rich Python terminal-formatting library from Textualize, running inside OpenCode, the open source terminal coding agent. OpenCode’s built-in DeepSeek list only ships with Flash, so the author added a small config file to get Pro into the model picker. For each test, the author built a separate clone of the repo, started a fresh session, and used identical prompts, recording time from OpenCode’s session timer and tokens, API requests, and cost from DeepSeek’s usage dashboard. The three tests were a bug fix, a feature build, and a performance optimization. The bug fix involved a user reporting that importing Rich crashes on macOS with Rich 15.0.0, Python 3.14, macOS on Apple Silicon, and a VS Code integrated terminal inside a Dropbox File Provider folder, where Rich calls os.getcwd() at import time and only guards against FileNotFoundError, but macOS File Provider folders make getc