bashkit
Benches

Latest benchmark snapshot

Static aggregate generated from repository result artifacts. Use the linked files for raw measurements and full eval traces.

Latest reports

Open Markdown reports

Benchmark folder Eval folder Criterion folder
bashkit-bench report 96 cases on vm-linux-x86_64 LLM eval report gpt-5.3-codex, 54/58 tasks Criterion report file-ops, 11 cases
Runtime snapshot

Latest benchmark categories

Browse benchmark runs
CategoryCasesLast run
startup Small commands where interpreter startup dominates runtime. 4 0.053 ms bash median: 1.662 ms
strings String expansion, pattern handling, and text manipulation. 8 0.057 ms bash median: 1.791 ms
variables Variable assignment, lookup, expansion, and environment handling. 8 0.058 ms bash median: 1.688 ms
arrays Indexed array reads, writes, expansion, and iteration. 6 0.059 ms bash median: 1.713 ms
subshell Command substitution and nested shell execution paths. 6 0.061 ms bash median: 3.143 ms
arithmetic Integer math, substitutions, and expression-heavy shell snippets. 6 0.062 ms bash median: 1.703 ms
pipes Pipeline construction, streaming, and command chaining. 6 0.065 ms bash median: 3.131 ms
control Conditionals, loops, case statements, and branching scripts. 9 0.076 ms bash median: 1.711 ms
Eval pressure

Lowest eval categories

Browse eval runs
Latest LLM eval 93% 54/58 tasks
CategoryPassedPass rate
system_info 1/2 50% tasks passed
file_operations 3/4 66.7% tasks passed
scripting 5/7 68.6% tasks passed
json_processing 8/8 100% tasks passed
data_transformation 6/6 100% tasks passed
complex_tasks 6/6 100% tasks passed
text_processing 6/6 100% tasks passed
pipelines 5/5 100% tasks passed