Kimi K2.7 Code Is Now Available in Puter.js

Puter.js now supports Kimi K2.7 Code, Moonshot AI's open-weight coding-agent model built for long-horizon, autonomous software engineering.

What is Kimi K2.7 Code?

Kimi K2.7 Code is a coding-focused open-weight model from Moonshot AI, released in June 2026. It shares the same 1-trillion-parameter mixture-of-experts architecture as Kimi K2.6, with 32 billion active parameters per token, but it is dedicated entirely to software engineering workloads.

Key highlights:

• 1T-parameter MoE architecture — trillion-parameter capacity with 32B active parameters per token for efficient sparse activation
• 262K context window — work across multi-file, repository-scale codebases and long agent traces in a single prompt
• Coding-first design — purpose-built for long-horizon, autonomous coding tasks and agentic workflows
• Always-on thinking mode — runs in thinking mode by default for sustained reasoning; non-thinking mode is not supported
• Open weights — model weights are published on Hugging Face under a Modified MIT license

What's New Compared to Kimi K2.6?

K2.7 Code narrows its focus to software engineering and improves both quality and efficiency:

• +21.8% on Kimi Code Bench v2 — 62.0 versus 50.9 for K2.6
• +11% on Program Bench — 53.6 versus 48.3 for K2.6
• +31.5% on MLS Bench Lite — 35.1 versus 26.7 for K2.6
• ~10% gains on agent execution benchmarks — stronger results on autonomous task benchmarks like MCP Atlas and MCP Mark Verified
• ~30% fewer reasoning tokens — reaches its answers with less thinking-token usage than K2.6
• Persistent reasoning context — preserves its thought process across interactions, useful for large projects that require maintaining long-term state

Examples

Agentic Coding

puter.ai.chat(
  "Refactor this Go service to use context cancellation and propagate errors cleanly across goroutines. Explain each change.",
  { model: 'moonshotai/kimi-k2.7-code' }
);

Repository-Scale Code Review

puter.ai.chat([
  { role: "user", content: largeCodebase + "\n\nIdentify architectural issues, dead code, and opportunities for simplification. Group findings by file." }
], { model: 'moonshotai/kimi-k2.7-code' });

Generating Code

puter.ai.chat(
  "Write a Python function that implements binary search on a sorted array, with tests.",
  { model: 'moonshotai/kimi-k2.7-code' }
);

Streaming

const response = await puter.ai.chat(
  "Design a fault-tolerant job queue in Rust with at-least-once delivery semantics.",
  { model: 'moonshotai/kimi-k2.7-code', 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 and no infrastructure setup. Start building with Kimi K2.7 Code immediately.

Learn more:

• Free, Unlimited Kimi API
• Puter.js AI Documentation