KAT Code Bench

Agents turn a natural-language requirement into a patch in a real repository drawn from Kuaishou internal development work. Each task pins its commit, environment and an executable verification entry point across 12 languages.

In-house benchmark described in the KAT-Coder-V2.5 tech report. It covers bug fixes, feature work, compatibility and cross-module edits.

Made by Kuaishou (KwaiKAT team) Released Jul 2026 Fixing real repositories Tasks kept private Last in a launch post Jul 2026 Last model added Jul 2026

Level: University Undergraduates and entry-level professionals
Scored: Checked automatically A program checks the result against an answer key, unit tests or the expected final state.
Answers: Agent task Many steps in a terminal, browser, desktop, app, game or robot.

Paper

Models scored in lab cards

Launch posts list it, from 1 lab

KAT-Coder-Pro V2.5 leads with 53.1%

Best score per model as labs report them in their own cards. Labs pick their own settings, so these compare less cleanly than independent runs.

Bars Table

1 KAT-Coder-Pro V2.5 53.1%
2 KAT-Coder-V2.5-Dev 46.2%

#	Model	Score	Setting	Date	Source
1	KAT-Coder-Pro V2.5	53.1%		Jul 2026	Model card
2	KAT-Coder-V2.5-Dev	46.2%		Jul 2026	Model card

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Reported by the lab in its model card

One lab put it in a launch post in the last six months

Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.

Kuaishou

Date	Launch post	Score	As the post writes it
23 Jul 2026	KAT-Coder-V2.5-Dev	46.2	KAT-Code-Bench
6 Jul 2026	KAT-Coder-Pro V2.5	53.1	KAT Code Bench

KAT-Coder-Pro V2.5 leads with 53.1%

Scores over time

One lab put it in a launch post in the last six months

More in Coding