Kwaipilot: KAT-Coder-Pro V1

Text input Text output Unavailable
Author's Description

KAT-Coder-Pro V1 is KwaiKAT's most advanced agentic coding model in the KAT-Coder series. Designed specifically for agentic coding tasks, it excels in real-world software engineering scenarios, achieving 73.4% solve rate on the SWE-Bench Verified benchmark. The model has been optimized for tool-use capability, multi-turn interaction, instruction following, generalization, and comprehensive capabilities through a multi-stage training process, including mid-training, supervised fine-tuning (SFT), reinforcement fine-tuning (RFT), and scalable agentic RL.

Key Specifications
Cost
$$$
Context
256K
Released
Nov 09, 2025
Speed
Ability
Reliability
Supported Parameters

This model supports the following parameters:

Tool Choice Tools Temperature Max Tokens Stop Top P Frequency Penalty
Features

This model supports the following features:

Tools
Performance Summary

Kwaipilot's KAT-Coder-Pro V1, created on November 9, 2025, is an advanced agentic coding model optimized for real-world software engineering tasks, achieving a 73.4% solve rate on SWE-Bench Verified. The model performs among the fastest models, ranking in the 65th percentile for speed, and offers competitive pricing, placing in the 69th percentile for cost-effectiveness. Notably, KAT-Coder-Pro V1 demonstrates exceptional reliability with a 100% success rate across all benchmarks, indicating minimal technical failures. Its strengths lie in General Knowledge and Ethics, both achieving 99.0% accuracy, and strong Reasoning capabilities at 84.0%. The model also shows solid performance in Mathematics (88.0%) and Coding (82.0%). A key area for improvement is its handling of Hallucinations, where it scored 82.0% accuracy, indicating a tendency to not always acknowledge uncertainty. Email Classification (95.0%) and Instruction Following (60.0%) also present opportunities for enhancement. Overall, KAT-Coder-Pro V1 is a robust agentic coding model with high reliability and strong performance in core knowledge and reasoning tasks, though its ability to acknowledge uncertainty could be refined.

Model Pricing

Current Pricing

Feature Price (per 1M tokens)
Prompt $0.207
Completion $0.828
Input Cache Read $0.0414

Price History

Available Endpoints
Provider Endpoint Name Context Length Pricing (Input) Pricing (Output)
StreamLake
StreamLake | kwaipilot/kat-coder-pro-v1 256K $0.207 / 1M tokens $0.828 / 1M tokens
AtlasCloud
AtlasCloud | kwaipilot/kat-coder-pro-v1 256K $0.207 / 1M tokens $0.828 / 1M tokens
Novita
Novita | kwaipilot/kat-coder-pro-v1 256K $0.207 / 1M tokens $0.828 / 1M tokens
Benchmark Results
Benchmark Category Reasoning Strategy Free Executions Accuracy Cost Duration
Other Models by kwaipilot