Moonshot
Kimi K3
kimi-k3
Total Context
1M
Max Output
131.1K
Released
N/A
Input$2.8571 / M tokens
Output$14.2857 / M tokens
Kimi K2.7 Code
kimi-k2.7-codeKimi K2.7 Code is a programming-oriented model optimized for software engineering scenarios. It helps developers generate, understand, and improve code, while handling tasks such as debugging, code review, project analysis, and automated development workflows. It is a strong choice for coding assistants, developer agents, and AI-powered programming tools.
Context Window
262.1K tokens
Maximum Output
262.1K tokens
Release Date
Jun 12, 2026
Modalities
Kimi K2.7 Code Pricing
| Input Price | Output Price | Cache Read |
|---|---|---|
| $0.9286/M | $3.8571/M | $0.1857/M |
Kimi K2.7 Code API Capabilities
Reasoning
Supported
Tool calling
Supported
Temperature parameter
Not supported
Attachments
Supported
Knowledge Base
2025-01-01
Endpoint Protocols
Completions APIMessages API
Kimi K2.7 Code Model Highlights
Kimi K2.7 Code is an open-weight multimodal coding model built for long-horizon software engineering, efficient reasoning and repository-scale context.
Long-Horizon Coding
Maintains instructions across extended coding sessions and improves end-to-end completion of complex, multi-file engineering tasks.
Reasoning Efficiency
Uses about 30% fewer thinking tokens on average than Kimi K2.6 while improving results on the cited coding evaluations.
Multimodal Code Reasoning
Its native multimodal architecture processes text, images and video alongside code within a 256K-token context.
Kimi K2.7 Code Use Cases
Kimi K2.7 Code suits repository-wide changes, extended debugging sessions and visually guided frontend development.
Repository-Scale Refactoring
Inspect a large codebase, trace dependencies and implement coordinated refactors across multiple files while preserving requirements.
Long-Running Debugging
Investigate defects over many iterations, retain prior findings and revise the implementation until tests and expected behavior align.
Visual Frontend Reconstruction
Interpret screenshots or video references and translate their layout, styling and interactions into working frontend code.
How to Use Kimi K2.7 Code via the TokenHub API
import OpenAI from "openai"
const client = new OpenAI({
apiKey: process.env.TOKENHUB_API_KEY,
baseURL: "https://us-api.tokenhub.com/v1",
})
const result = await client.chat.completions.create({})
console.log(result.choices[0]?.message?.content)Kimi K2.7 Code Benchmarks
Kimi K2.7 Code
| Index score | ||
|---|---|---|
| Artificial Analysis Intelligence Index | Artificial Analysis broad capability aggregate | 41.9 |
| Artificial Analysis Coding Index | Artificial Analysis software task aggregate | 60.8 |
| Knowledge & Reasoning | ||
| GPQA | Advanced science problem solving | 89.6% |
| HLE | Broad expert-level exam set | 32.8% |
| Coding & Engineering | ||
| SciCode | Scientific coding challenges | 47.5% |
| Terminal-Bench Hard | Hard terminal task execution | 44.7% |
| Instruction Following & Agent Tasks | ||
| IFBench | Prompt constraint adherence | 63.1% |
| AA-LCR | Long-context reasoning | 66.3% |
| τ²-Bench | Agent workflow tasks | 90.1% |
Metrics sourced from Artificial Analysis
Kimi K2.7 Code FAQs
Common questions about using Kimi K2.7 Code on TokenHub.
What is Kimi K2.7 Code?+
Kimi K2.7 Code is a coding-focused agentic model designed for software development and long-horizon programming tasks.
What is Kimi K2.7 Code best for?+
It is well suited to repository navigation, code generation, refactoring, debugging, testing, and tool-using coding agents.
What are its main strengths and tradeoffs?+
It offers strong coding-agent behavior and a 256K context window, but reasoning can increase latency and cost; review important changes before merging.
How should I call Kimi K2.7 Code through TokenHub?+
Select this model in TokenHub and use the provider's documented API-compatible settings; technical integrations may require the model ID kimi-for-coding.
Who should choose Kimi K2.7 Code instead of Kimi K2.6 or Kimi K3?+
Choose it when software engineering is the priority and you want a stable coding-focused workflow without needing Kimi K3's million-token context.
Ready to use Kimi K2.7 Code?
Use one API key to access Kimi K2.7 Code and more AI models through TokenHub.
Kimi K2.7 Code Media, Demos and Reviews
Public launch posts, hands-on tests and community discussion about Kimi K2.7 Code.
X (Twitter)
Mark Kretschmann ·
Kimi K2.7 Code lands at #2 on ErdosBench, right behind Claude Fable 5 max. Very strong result: 13/14 coverage, 5 fully accepted solves, 5 accepted partials, 0 major false partials, 0 unsafe partials. It even matches the top model on the five accepted solved/settled results. The https://t.co/AceDr1WdXN
https://pbs.twimg.com/media/HKxwkWmXcAAF61V.jpgView originalNoctus ·
Kimi K2.7 is absolutely incredible Some of the things it pulls off in a single shot are absurd. https://t.co/3lmSiPrRid
https://pbs.twimg.com/media/HKwcvqebUAAcNUc.jpghttps://pbs.twimg.com/amplify_video_thumb/2066057981595381760/img/no8GVj9APko6HBpt.jpgView originalstevibe ·
I gave Kimi K2.6 and K2.7 Code the exact same prompt to animate a letter burning to ash. Pure HTML canvas, zero libraries. https://t.co/MgLx9XfMrR
https://pbs.twimg.com/amplify_video_thumb/2065463846324842496/img/DlbNXkT5ZgvGa8ik.jpgView originalYouTube
Kimi K2.7 Code: The first open-weight model in GitHub Copilot
GitHub
View originalKimi K2.7 Code Is HERE – Is THIS the Best Open Coding Model Yet?
Bijan Bowen
View originalI Tested NEW Kimi-K2.7-Code with 20 Prompts
AI Coding Daily
View original