PDB-Single - Edit Precision: leaderboard

The Precise Debugging Benchmark (PDB, 2026) turns coding datasets into debugging tasks with verified injected bugs and checks whether a model fixes them with minimal targeted edits instead of regenerating the program. This board is edit-level precision in percent on the PDB-Single-Full pool of programs with one injected single-line bug; higher is better.

Metric: Edit-level precision (%): share of a model's predicted edits that match an essential ground-truth fix, on the PDB-Single-Full set of single-bug programs; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 8 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1Claude Sonnet 4.5Thinking78.1#261
2Gemini 2.5 Pro77.9#276
3Qwen 3 Coder 480B A35B73.5#709
4Kimi K265.8#443
5Grok Code Fast 163.8#703
6Kimi K2 (Thinking)61.3
7DeepSeek V3.258.6#419
8GPT-5.1 Codex50.3#295

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=pdb-single-edit-precision · How It Works · Data refreshed daily, snapshot 2026-11-21.