KAT-Coder-Pro V1

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.

Share

Model details

Context window256,000 tokens
Max completion size73 tokens
Prompt cost / 1K tokens$0.000000207
Completion cost / 1K tokens$0.000000828
Accepts
Produces

Benchmark performance

Overall

68
score
17th
placement

Cost

97
score
4th
placement

Logic

80
score
7th
placement

Speed

81
score
19th
placement

Scoring

8
score
25th
placement

Tool Use

45
score
5th
placement

Hallucination

89
score
5th
placement

Classification

19
score
4th
placement

Structured Output

83
score
3rd
placement

Pricing

Usage pricing
Prompt
$0.000000207
Completion
$0.000000828
Request
FREE
Image
FREE
Web Search
FREE
Internal Reasoning
FREE
Input Cache Read
FREE

Best Overall scoring LLMs

88
score
1st
placement
86
score
2nd
placement
84
score
3rd
placement
82
score
4th
placement
82
score
4th
placement
80
score
5th
placement