AI
MarkTechPost AI模型更新

Moonshot AI 推出 Kimi K2.7-Code：程式碼模型，在 Kimi Code Bench v2 上較 K2.6 提升 21.8%

2026年6月13日 04:57

重點摘要

本週，Moonshot AI 推出 Kimi K2.7-Code。這是一款專注於程式碼的智慧代理模型，採用修改後的 MIT 授權條款於 Hugging Face 釋出模型權重，並可透過 Kimi API 與 Kimi Code 使用。K2.7-Code 針對長週期軟體工程任務設計，而非一般對話用途，能進行規劃、編輯、執行工具與跨多步驟除錯。Moonshot 亦圍繞此模型提供訂閱制的程式開發平臺。K2.7-Code 為專家混合模型，總參數達 1 兆，每個 Token 啟動 32B 參數，設計採用 384 位專家（每個 Token 選取 8 位加上 1 位共享專家），共 61 層（含 1 個稠密層）。注意力機制使用 MLA，前饋路徑採用 SwiGLU，另加入 MoonViT 視覺編碼器（4 億參數）以支援影像與影片輸入。模型原生支援……（原文截斷）

站內 AI 整理稿

This week, Moonshot AI released Kimi K2.7-Code. It is a coding-focused, agentic model. The model weights ship on Hugging Face under a Modified MIT license. You can also reach it through the Kimi API and Kimi Code. K2.7-Code targets long-horizon software engineering, not general chat. It plans, edits, runs tools, and debugs across many steps. Moonshot pairs the model with a subscription coding platform around it. Kimi K2.7-Code K2.7-Code is a Mixture-of-Experts model. It holds 1T total parameters and activates 32B per token. The design uses 384 experts, with 8 selected per token and 1 shared. It has 61 layers, including 1 dense layer. Attention uses MLA, and the feed-forward path uses SwiGLU. A MoonViT vision encoder adds 400M parameters for image and video input. The model ships with native INT4 quantization. The context window is 256K tokens (262,144). Two constraints matters: Thinking mode is mandatory; disabling it returns an API error. Sampling is fixed: temperature 1.0, top_p 0.95, n 1, penalties 0.0. Default max output is 32,768 tokens. You can self-host with vLLM, SGLang, or KTransformers. The Hugging Face repository is large, roughly 595 GB on disk. This is a server-class deployment target, not a laptop model. Benchmark Moonshot team published six benchmark rows. They compare K2.7-Code against K2.6, GPT-5.5, and Claude Opus 4.8. K2.7-Code beats K2.6 on every row. The largest coding jump is Kimi Code Bench v2, from 50.9 to 62.0. BenchmarkKimi K2.6Kimi K2.7-CodeGPT-5.5Claude Opus 4.8K2.7 vs K2.6Kimi Code Bench v250.962.069.067.4+21.8%Program Bench48.353.669.163.8+11.0%MLS Bench Lite26.735.135.542.8+31.5%Kimi Claw 24/7 Bench42.946.952.850.4+9.3%MCP Atlas69.476.079.481.3+9.5%MCP Mark Verified72.881.192.976.4+11.4% K2.7-Code does beat Opus 4.8 on MCP Mark Verified, 81.1 versus 76.4. It also lands close to GPT-5.5 on MLS Bench Lite. K2.7-Code ran in Kimi Code CLI, GPT-5.5 in Codex xhigh, and Opus 4.8 in Claude Code xhigh. Reasoning-Token Efficiency: A Cost Claim, Not Just Quality Moonshot team reports about 30% lower reasoning-token usage than K2.6. It frames this as ‘less overthinking.’ Reasoning tokens bill as output tokens on most price cards. Agentic coding runs hundreds or thousands of steps. Each plan, retry, and verification pays the thinking cost again. A 30% cut compounds across a long run. The effect lands in three places at once. First, lower output-token cost per task. Second, faster steps, which helps interactive CLI sessions. Third, more steps before hitting context limits. Use Cases With Examples Repo-scale refactors are the main use case. Point the agent at a failing test suite. It reads files, edits across modules, then reruns tests until green. Code review is a second fit. Feed a pull request diff and ask for risk analysis. The 256K window holds large diffs, logs, and related files together. MCP tool-use workflows are a third fit. K2.7-Code scored 81.1 on MCP Mark Verified. That suite tests correct tool invocation through the Model Context Protocol. Think CI checks, ticket updates, and file edits in one loop. Long-context analysis is a fourth fit. The model accepts text, image, and video input. Documentation, screenshots, and a recorded repro can share one prompt. Marktechpost’s Interactive Explorer #mtp-k27-demo *{box-sizing:border-box!important;margin:0;padding:0} #mtp-k27-demo{ background:#111!important;color:#e7e7e7!important; font-family:-apple-system,BlinkMacSystemFont,"Segoe UI",Roboto,Helvetica,Arial,sans-serif!important; border:1px solid #222!important;border-radius:14px!important; padding:22px!important;max-width:920px;margin:0 auto;line-height:1.5; } #mtp-k27-demo .k27-head{display:flex;align-items:center;gap:12px;flex-wrap:wrap;margin-bottom:4px} #mtp-k27-demo .k27-dot{width:11px;height:11px;border-radius:50%;background:#76B900;box-shadow:0 0 10px #76B900} #mtp-k27-demo h2{font-size:20px;color:#fff!important;font-weight:700;letter-spacing:-.2px} #mtp-k27-demo .k27-sub{color:#9aa0a6!important;font-size:13px;margin:2px 0 16px} #mtp-k27-demo .k27-tabs{display:flex;gap:6px;flex-wrap:wrap;margin-bottom:18px} #mtp-k27-demo .k27-tab{ background:#181818!important;color:#cfcfcf!important;border:1px solid #2a2a2a!important; padding:9px 14px;border-radius:9px;cursor:pointer;font-size:13px;font-weight:600;transition:.15s } #mtp-k27-demo .k27-tab:hover{border-color:#76B900!important} #mtp-k27-demo .k27-tab.on{background:#76B900!important;color:#0a0a0a!important;border-color:#76B900!important} #mtp-k27-demo .k27-panel{display:none} #mtp-k27-demo .k27-panel.on{display:block} #mtp-k27-demo .k27-pills{display:flex;gap:8px;flex-wrap:wrap;margin-bottom:16px} #mtp-k27-demo .k27-pill{ display:flex;align-items:center;gap:7px;background:#181818!important;border:1px solid #2a2a2a!important; padding:7px 11px;border-radius:20px;cursor:pointer;font-size:12px;color:#bdbdbd!important;user-select:none } #mtp-k27-demo .k27-pill .sw{width:11px;height:11px;border-radius:3px} #mtp-k27-demo .k27-pill.off{opacity:.38} #mtp-k27-demo .k27-bench{margin-bottom:16px} #mtp-k27-demo .k27-bname{font-size:13px;color:#dcdcdc!important;font-weight:600;margin-bottom:7px} #mtp-k27-demo .k27-row{display:flex;align-items:center;gap:10px;margin-bottom:6px} #mtp-k27-demo .k27-mlabel{width:120px;min-width:120px;font-size:11px;color:#9aa0a6!important} #mtp-k27-demo .k27-track{flex:1;background:#1c1c1c!important;border-radius:6px;height:22px;overflow:hidden} #mtp-k27-demo .k27-fill{height:100%;border-radius:6px;width:0;transition:width .7s cubic-bezier(.22,1,.36,1); display:flex;align-items:center;justify-content:flex-end;padding-right:8px;font-size:11px;font-weight:700;color:#0a0a0a} #mtp-k27-demo .k27-note{font-size:11px;color:#7c8085!important;margin-top:10px;border-top:1px solid #222!important;padding-top:10px} #mtp-k27-demo .k27-calc{display:grid;grid-template-columns:1fr 1fr;gap:16px} #mtp-k27-demo .k27-field{margin-bottom:12px} #mtp-k27-demo .k27-field label{display:block;font-size:12px;color:#bdbdbd!important;margin-bottom:6px} #mtp-k27-demo .k27-field .val{color:#76B900!important;font-weight:700} #mtp-k27-demo input[type=range]{width:100%;accent-color:#76B900;cursor:pointer} #mtp-k27-demo .k27-out{background:#161616!important;border:1px solid #262626!important;border-radius:11px;padding:16px} #mtp-k27-demo .k27-line{display:flex;justify-content:space-between;font-size:13px;padding:7px 0;border-bottom:1px dashed #262626!important} #mtp-k27-demo .k27-line:last-child{border-bottom:0} #mtp-k27-demo .k27-line b{color:#fff!important} #mtp-k27-demo .k27-total{font-size:22px;color:#76B900!important;font-weight:800;margin-top:6px} #mtp-k27-demo .k27-save{background:#13210a!important;border:1px solid #2f4d14!important;border-radius:9px; padding:10px 12px;font-size:12px;color:#aadd72!important;margin-top:12px} #mtp-k27-demo .k27-specs{display:grid;grid-template-columns:1fr 1fr;gap:10px} #mtp-k27-demo .k27-spec{background:#161616!important;border:1px solid #242424!important;border-radius:10px;padding:12px} #mtp-k27-demo .k27-spec .l{font-size:11px;color:#9aa0a6!important;text-transform:uppercase;letter-spacing:.4px} #mtp-k27-demo .k27-spec .v{font-size:14px;color:#fff!important;font-weight:600;margin-top:4px} @media (max-width:640px){ #mtp-k27-demo{padding:16px!important} #mtp-k27-demo .k27-calc{grid-template-columns:1fr} #mtp-k27-demo .k27-specs{grid-template-columns:1fr} #mtp-k27-demo .k27-mlabel{width:84px;min-width:84px} } Kimi K2.7-Code — Interactive Explorer Company-reported benchmarks and official API pricing. Released June 12, 2026. Verified June 12, 2026. Benchmarks Cost Calculator Specs Source: Moonshot AI Kimi K2.7-Code model card. K2.7-Code ran in Kimi Code CLI; GPT-5.5 in Codex xhigh; Claude Opus 4.8 in Claude Code xhigh. First-party numbers, not an independent leaderboard. Input tokens / run: 50,000 Output tokens / run: 8,000 Cache hit rate: 50% Runs / month: 1,000 Reasoning share of output: 40% Input cost$0.00 Output cost$0.00 Est. monthly total$0.00 $0.00 Rat

原始來源：MarkTechPost AI ↗
查看原始來源

Related

相關文章

MarkTechPost AI模型更新

Liquid AI Introduces LFM2.5-Embedding-350M and LFM2.5-ColBERT-350M: Dense Bi-Encoder and Late-Interaction Models for Fast Multilingual Search Across 11 Languages

This week, Liquid AI released two new retrieval models. They are LFM2.5-ColBERT-350M and LFM2.5-Embedding-350M. Both hold 350M parameters. Both are the first bidirectional members of the LFM family. They build on LFM2.5-350M-Base, released in March. The pair targets fast multilingual and cross-lingual search across 11 languages. Their footprint is small enough to run almost anywhere. Both are available now on Hugging Face under the LFM Open License v1.0. LFM2.5 Retrievers The two models share one backbone but represent text differently. LFM2.5-Embedding-350M is a dense bi-encoder. It turns each document into a single vector. Pick it when you want the fastest search and the smallest, cheapest index. LFM2.5-ColBERT-350M is a late-interaction model. It converts each token into a vector rather

21 分鐘前
MarkTechPost AI模型更新

Perplexity Launches Brain, a Self-Improving Memory System That Builds a Context Graph of an Agent’s Work and Learns Overnight

Most AI memory remembers the user. It stores your preferences, your tastes, and your role. Perplexity is taking a different path. Today, Perplexity launched Brain, a self-improving memory system for its agent product, Computer. Brain does not focus on remembering you. It remembers what the agent did. That reframes what memory in AI is for. What is Perplexity‘s Brain Brain is a self-improving memory system. It builds a context graph of the work Computer performs. At set intervals, such as overnight, Brain reviews that graph. It then teaches itself how to do the work better. The idea is straightforward. The more work you do, the more efficient Brain makes your Computer. Brain is rolling out today to Perplexity Max and Enterprise Max subscribers in Research Preview. Two Axes of AI Memory Perp

14 小時前
36氪模型更新

智譜新高，MiniMax承壓，“大模型雙雄”命運殊途

這篇消息聚焦「智譜新高，MiniMax承壓，“大模型雙雄”命運殊途」。原始導語提到：大模型在被市場重新定價 從 AI 情報角度來看，這類內容值得關注其背後的技術進展、產品落地、產業競爭與後續市場影響。

16 小時前