Benchmarks | GitBench

blame_forensics Avg: 85.5%

Best: 2 models · 100.0% Worst: liquid/lfm-2-24b-a2b · 16.7%

branch_cleanup Avg: 81.9%

Best: arcee-ai/trinity-large-thinking · 100.0% Worst: 2 models · 41.7%

cherry_pick Avg: 73.0%

Best: arcee-ai/trinity-large-thinking · 89.6% Worst: liquid/lfm-2-24b-a2b · 33.3%

commit_messages Avg: 91.5%

Best: 17 models · 100.0% Worst: moonshotai/kimi-k3 · 3.3%

commit_squash Avg: 81.1%

Best: 3 models · 100.0% Worst: liquid/lfm-2-24b-a2b · 8.3%

git_bisect Avg: 94.5%

Best: 32 models · 100.0% Worst: qwen/qwen3.6-flash · 49.2%

git_clean Avg: 85.3%

Best: 7 models · 100.0% Worst: liquid/lfm-2-24b-a2b · 25.0%

git_grep Avg: 88.9%

Best: 5 models · 100.0% Worst: qwen/qwen3.6-flash · 45.8%

git_log_format Avg: 92.9%

Best: 22 models · 100.0% Worst: qwen/qwen3.6-flash · 49.2%

git_show Avg: 90.4%

Best: 8 models · 100.0% Worst: qwen/qwen3.6-flash · 45.8%

merge_conflicts Avg: 75.1%

Best: google/gemini-3.6-flash · 91.7% Worst: liquid/lfm-2-24b-a2b · 33.3%

rebase Avg: 74.5%

Best: moonshotai/kimi-k3 · 90.0% Worst: qwen/qwen3.6-flash · 38.3%

reflog Avg: 89.9%

Best: 14 models · 100.0% Worst: 3 models · 50.0%

stash_recovery Avg: 94.8%

Best: 36 models · 100.0% Worst: 3 models · 50.0%

submodule_usage Avg: 87.9%

Best: 3 models · 100.0% Worst: qwen/qwen3.7-flash · 46.5%

tag_management Avg: 89.3%

Best: 12 models · 100.0% Worst: 2 models · 49.2%

worktree_usage Avg: 81.7%

Best: x-ai/grok-4.3 · 98.6% Worst: liquid/lfm-2-24b-a2b · 8.3%