Kimi K2.7 Code
द्वारा Moonshot · रिलीज़ 2026
Moonshot AI's frontier-scale open-source 1T-parameter model tuned for agentic coding. 262K context window, multi-turn tool calling, reasoning, vision inputs, and structured outputs for agentic workloads.
Kimi K2.7 Code
द्वारा संचालित Moonshot · Mixture-of-Experts (1T params, agentic coding)
कॉन्टेक्स्ट विंडो
262K
पैरामीटर
1T (MoE)
अधिकतम आउटपुट
16K
श्रेणी
LLM चैट
अवलोकन
Kimi K2.7 Code is Moonshot AI's frontier-scale open-source model — a 1-trillion-parameter mixture-of-experts architecture purpose-built for agentic coding workloads. It pairs a 262,144-token context window with multi-turn tool calling, native reasoning, vision inputs, and structured outputs, making it well-suited for long-horizon software-engineering tasks that span large codebases and many tool invocations.
The model is designed for autonomous coding agents: it can plan a change across many files, call tools to read and edit code, run tests, and iterate — all while keeping the full project context in its very large window. Vision input lets it work from screenshots, diagrams, and design mockups, and structured output keeps tool-call payloads reliable.
Served through CallMissed on Cloudflare Workers AI, K2.7 Code is OpenAI-compatible (`/v1/chat/completions`) with streaming, full tool calling, and a `reasoning_effort` thinking toggle.
प्राइसिंग
| मेट्रिक | कीमत |
|---|---|
| इनपुट /1M tokens | ₹128.0000 |
| आउटपुट /1M tokens | ₹540.0000 |
1 क्रेडिट = ₹1 = $0.01 USD। कीमतें प्रोवाइडर से दिखाई गई हैं; CallMissed ~35% मार्कअप के साथ पास-थ्रू करता है।
मुख्य बातें
- Frontier-scale 1T-parameter MoE tuned for agentic coding
- 262K-token context for whole-codebase reasoning
- Multi-turn tool calling + structured outputs
- Vision inputs (screenshots, mockups, diagrams)
बेंचमार्क
| बेंचमार्क | स्कोर |
|---|---|
| SWE-bench | SOTA-class |
| LiveCodeBench | Top-tier |
तकनीकी विवरण
- Frontier-scale 1T-parameter mixture-of-experts architecture
- Context window: 262,144 tokens
- Multi-turn function/tool calling
- Native reasoning with reasoning_effort thinking toggle
- Vision inputs supported
- Structured outputs for agentic workflows
- OpenAI-compatible; served via Cloudflare Workers AI on the CallMissed gateway
ताकतें
- Massive 262K context keeps whole codebases in view
- Built for multi-turn tool-using coding agents
- Vision + reasoning + structured outputs in one model
सीमाएं
- Frontier-scale model — higher latency than small fast tiers
- Newer release with limited production track record
उपयोग के मामले
API उदाहरण
curl https://api.callmissed.com/v1/chat/completions \
-H "Authorization: Bearer cm_YOUR_KEY" \
-d '{"model": "kimi-k2.7-code", "messages": [{"role": "user", "content": "Refactor this module and add tests"}]}'एंडपॉइंट: POST /v1/chat/completions · मॉडल ID: kimi-k2.7-code
Kimi K2.7 Code अभी आज़माएं
साइनअप पर 1000 फ्री API क्रेडिट पाएं। कोई क्रेडिट कार्ड ज़रूरी नहीं।