KwaiKAT Team Releases KAT-Coder-V2.5: An Agentic Coding Model Trained on 100,000+ Verifiable Repository Environments
MarkTechPost Michal Sutter
Kuaishou's KwaiKAT team released KAT-Coder-V2.5, an AI coding model trained on over 100,000 verifiable repository environments where it learns to operate inside real, executable codebases rather than generate isolated code snippets. The model was trained using 100,000+ environments spanning 12 languages built by AutoBuilder, which improved environment construction success from 16.5% to 57.2%, and infrastructure fixes reduced sandbox-induced training failures from 16% to below 2%. KAT-Coder-V2.5 now ranks first on PinchBench (94.9) and second on SWE-Bench Pro (65.2), with an open-weight variant released on Hugging Face under Apache 2.0.
Why it matters
The KwaiKAT Team at Kuaishou has published the KAT-Coder-V2.5 technical report, arguing that agentic coding capability is bottlenecked by training infrastructure rather than model scale. AutoBuilder raised environment construction success from 16.5% to 57.2%, producing over 100,000 verifiable environments across 12 languages, while a sandbox audit cut RL feedback errors from roughly 16% to below 2%. The post KwaiKAT Team Releases KAT-Coder-V2.5: An Agentic Coding Model Trained on 100,000+ Verifiable Repository Environments appeared first on MarkTechPost.
