KAT-Coder-Pro V2 is theKwaiKAT’s Latest Flagship Agentic Coding Model. Engineered for multi-scaffold generalization, the model is natively compatible with over 10 mainstream AI coding tools—such as Claude Code, Cline, Kilo, and OpenCode, offering unparalleled flexibility. Through dedicated full-pipeline optimization for OpenClaw, KAT-Coder-Pro V2 is trained from the ground up to master complex, real-world application workflows with ease. Beyond logic, KAT-Coder-Pro V2 achieves a breakthrough in frontend aesthetic generation. In Landing Page and PPT scenarios, the model delivers a paradigm shift in user experience: No Structured Spec Needed: Users no longer need to provide rigid design specifications. Simply describe what you want in plain language to receive production-grade, high-quality output that rivals structured design inputs. Mass Market Expansion: This evolution expands the model's service boundary from the 1% of power users to hundreds of millions of ordinary users, truly democratizing professional-grade creation.
Providers
Route requests across multiple providers. Copy a provider slug to set your preference.
Uptime
24hoursDirect request success rate on AI Gateway and per-provider.
Throughput
24hoursP50 throughput on live AI Gateway traffic, in tokens per second (TPS).
Latency
24hoursP50 time to first token (TTFT) on live AI Gateway traffic, in milliseconds.
Activity
Token volume and request traffic to this model over time.
Token volume broken down by the provider that served each request. Larger segments indicate providers handling more of this model's traffic.
Benchmarks
Scores on standardized evaluations. Higher percentages are better — and rank percentile shows
Metrics sourced fromArtificial Analysis
Artificial Analysis
Apps
Public apps that send the most traffic to this model. Good signal for what real production workloads look like — and a hint at which use cases this model is best suited for. View All