Back to Models
Kwaipilot: KAT-Coder-Air V2.5
Kwaipilottext8.4 / 10 Overall Rating
Kwaipilot KAT-Coder-Air V2.5 is an agentic coding model built for autonomous software issue resolution and workflow automation. It provides a 256K token context window with an exceptionally high 80K max output limit at a fraction of the price of Claude 3.5 Sonnet and GPT-4o ($0.15/1M input, $0.60/1M output). It targets automated long-context code generation and codebase refactoring.
Context Window
256K
Knowledge Cutoff
2025
Max Output
16K
License
Proprietary
Where it Excels (Pros)
- 256K input context window
- 80K maximum output tokens
- Extremely cheap API pricing
- Optimized for agentic coding
Limitations (Cons)
- Unproven general reasoning performance
- Lacks multimodal input capabilities
Benchmark Breakdown
Reasoning (MMLU)82%
Coding (HumanEval)78%
Mathematics (MATH)76%
Long-Context Retrieval95%
Pricing Matrix
Input cost / 1M tokens$0.15
Output cost / 1M tokens$0.60
Compare Kwaipilot: KAT-Coder-Air V2.5 with other frontier models: