Kwaipilot KAT-Coder V2.5 Models Are Now Available in Puter.js

We just added the KAT-Coder V2.5 models from Kwaipilot (Kuaishou's AI research division) to Puter.js, KAT-Coder-Pro V2.5 and KAT-Coder-Air V2.5. Both are agentic coding models with a 256K token context window and up to 80K output tokens. Add them to your application for free without any API keys.

What is KAT-Coder-Pro V2.5?

KAT-Coder-Pro V2.5 is the successor to KAT-Coder-Pro V2, trained through large-scale agentic reinforcement learning in reconstructed, verifiable repository environments. It's built to take on an entire issue or business workflow autonomously, locating the problem, making changes, and completing the task inside a real repository. Key highlights:

• SWE-Bench Pro: 65.2, which Kwaipilot reports as second only to Claude Opus 4.8 (69.2)
• PinchBench: 94.9, the best agentic tool-use score among the models Kwaipilot tested
• 256K token context window with up to 80K output tokens
• $0.74 / $2.96 per million input/output tokens

What is KAT-Coder-Air V2.5?

KAT-Coder-Air V2.5 is the faster, lower-cost sibling, sharing Pro's 256K context window and 80K max output. It's built for agent and tool-use workflows inside real, executable repositories, handling issue localization, code modification, and test execution as part of an end-to-end development loop, and it integrates into toolchains such as Claude Code and OpenHands. Key highlights:

• $0.15 / $0.60 per million input/output tokens, roughly a fifth of Pro's price
• Function calling, tool use, and structured JSON output for agent workflows
• Prompt caching to cut costs on repeated context
• Built for rapid iteration, aimed at high-volume coding agents where speed and cost per request matter

Examples

Agentic coding with KAT-Coder-Pro V2.5

puter.ai.chat(
    "Implement an Express rate-limiter middleware with a token bucket algorithm, then write unit tests covering burst traffic and refill timing.",
    { model: 'kwaipilot/kat-coder-pro-v2.5' }
);

High-volume tasks with KAT-Coder-Air V2.5

puter.ai.chat(
    "Write a TypeScript utility that deep-merges two objects, with type definitions and edge-case handling.",
    { model: 'kwaipilot/kat-coder-air-v2.5' }
);

Streaming

const response = await puter.ai.chat(
    "Refactor this codebase to extract the payment logic into its own module. Walk through the changes step by step.",
    { model: 'kwaipilot/kat-coder-pro-v2.5', stream: true }
);

for await (const part of response) {
    puter.print(part?.text);
}

Get Started Now

Just add one library to your project:

// npm install @heyputer/puter.js
import { puter } from '@heyputer/puter.js';

Or add one script tag to your HTML:

<script src="https://js.puter.com/v2/"></script>

No API keys needed. Start building with the KAT-Coder V2.5 models immediately.

Learn more:

• Free, Unlimited Kwaipilot Kat API
• Puter.js AI Documentation