ForgeCode vs Claude Code: which AI coding agent actually wins?

The author was frustrated by the speed of Claude Code, experiencing significant delays on even simple tasks. They explored alternatives, discovering ForgeCode, an open-source model-agnostic agent harness. ForgeCode, particularly with Opus 4.6, outperformed Claude Code on speed, though GPT 5.4 was unstable. ForgeCode's success stems from optimizations like its context engine, but its ecosystem is less developed than Claude Code's. Benchmarks revealed ForgeCode's strong performance on its own tbench.ai benchmark but a smaller performance gap on independent benchmarks like SWE-bench. The author ultimately adopted a hybrid approach, using Claude Code for its broader features and ForgeCode for speed-sensitive tasks. ForgeCode's speed gains with Opus are undeniable. ForgeCode offers a fast core engine, but lacks features such as hooks and auto-memory found in Claude Code. The article provides insights for choosing between the agents as the author uses the tools. The author is using multiple tools to address the speed gains of ForgeCode. The author considers ForgeCode's performance to be good, but with limitations.