Atlarix vs opencode on Terminal-Bench 2.0 — same model, only the harness changes (k=1, receipts included)

Short summary

Benchmark compares Atlarix (agent workstation) vs opencode harnesses on Terminal-Bench 2.0 using the same model (minimax-m3). Atlarix resolved 42/89, opencode 39/89—a 3-task gap within k=1 noise, showing harness design doesn't bottleneck model performance. Both hit timeout limits on ~25% of tasks; raw results published for reproducibility.

• •Atlarix (42/89) and opencode (39/89) show harness design doesn't bottleneck model performance
• •3-task difference is within k=1 noise; author honestly avoids false victory claims
• •Timeout limits (~25% of tasks) constrain both equally; raw benchmark data published for verification

Generated with AI, which can make mistakes.