RTK reports token savings, but our cost benchmarks disagree

RTK compresses terminal output by stripping nonessential fields, but benchmark tests on Terminal-Bench 2.1 show it yields mixed cost results: Claude Code with Fable 5.0 becomes about 3% cheaper while DeepSeek V4 Pro becomes 7-18% more expensive, with lower pass rates. The apparent token savings reported by RTK (up to 89% reduction) do not translate into actual monetary savings because the tool alters subsequent model turns and can increase overall token consumption.