Skip to content

Moonshot AI · Chat

Kimi K2.7 Code

Moonshot AI 🇨🇳ChatKimi familyOpen weights. One of its sources says the weights are publicly available.The maker currently plans to deprecate it within the next 180 days (16 Apr 2027). Plans can change.
From Empiriolabs's listing:
Kimi K2.7 Code is Moonshot's trillion-parameter agentic coding model with 256K context, always-on reasoning, and text, image, and video inputs.
llminfo IQ
126.9
Max context
271K
Listings
135

Evidence and sources

International API storefront is Singapore-based; principal operating entity 北京月之暗面科技有限公司 (Beijing).

🤗 moonshotai/Kimi-K2.7-Code Model architecture, as its sources state it llminfo IQ 126.9, ranked among chat models. From 102 benchmarks. Plenty of data behind it.How llminfo IQ is computedCoding IQ 104.5, 95% interval 96–113. From 21 benchmark families across 11 sources. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.How Coding IQ is computedPoor value (43.6). Costs more per task than its benchmark results predict. From 21 benchmarks with measured task costs. Plenty of data behind it.How Value for money worksYap score: tokens used per coding task vs. models with the same resultsChatty (62 ±9). Generates more tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.Median 52,402 output tokens per task at default effort.How the Yap score worksReleased 11 Jun 2026Maker HQ: CN (source)
All rankings: #68 on NanoGPT recently #109 on OpenRouter today #116 on OpenRouter this week #124 on OpenRouter this monthSources: LMArena data, EmpirioLabs data, LiteLLM data, Models.dev data, Morpheus Inference Marketplace data, Hugging FacePrimary data from LMArena from EmpirioLabs: primary mode, description from LiteLLM: deprecation date from Models.dev: weights availability from Morpheus Inference Marketplace: architecture from Hugging Face: release date
Tags from its sources:instructtoolsreasoningvisiondocumentopen-sourcemoelong-contextcoding

Advantages and trade-offs

Compared with 10 chat models, picked mainly by popularity and IQ

Advantages

Open weights available

The weights are public. 3 of the 9 peers that say either way have open weights.

Trade-offs

Deprecation scheduled

The maker plans to deprecate it on 16 Apr 2027, within the next 180 days.

We only make a claim when there's enough data on models of the same type. Updated 22 Nov, 04:34 UTC.

Input & output

as stated by the model's providers
Text
Images
Video
Files
Text

Capabilities

Supported operations and provider-stated features

VisionReasoningFunction callingStructured outputPrompt cachingWeb search

Listings

 Moonshot AIkimi-k2.7-code · int4Sources: GenAI Prices data, LiteLLM data, Models.dev data262K$0.95cache read $0.19$4.00——
 Routing Runkimi-k2.7-codeSource: Models.dev data200K$0.275$1.10——
 Kenarikimi-k2-7-codeSource: Models.dev data262K$0.5714cache read $0.1143$2.8571——
 Hugging FaceInference by Coreweavemoonshotai/Kimi-K2.7-CodeSource: Hugging Face Inference Providers data262K$0.60$3.00——
 Kilomoonshotai/kimi-k2.7-codeSources: Kilo AI Gateway data, Models.dev data262K$0.60cache read $0.12$3.00——
 NanoGPTmoonshotai/kimi-k2.7-codeSources: Models.dev data, NanoGPT262K$0.60cache read $0.12$3.00——
 Nous Portalmoonshotai/kimi-k2.7-codeSource: Nous Portal Inference Gateway data262K$0.60cache read $0.12$3.00——
 OpenRouterInference by Coreweavemoonshotai/kimi-k2.7-code · int4Source: OpenRouter data docs262K$0.60cache read $0.12$3.0022804ms
 Weights & Biasesmoonshotai/Kimi-K2.7-CodeSources: LiteLLM data, Models.dev data, Weights & Biases Inference data262K$0.60cache read $0.12$3.00——
 Inceptronmoonshotai/Kimi-K2.7-Code · int4Sources: Inceptron data, Models.dev data262K$0.61cache read $0.17$3.55——

12 listings without a price are hidden.

Included in plans

Umans Code ProumansSubscription$20/moslows down past the allowancePlan page
Umans Code MaxumansSubscription$50/moslows down past the allowancePlan page

Benchmarks

Named effort levels run from lowest to highest. Select a result to see its source details.

Agentic

Design Arena EloSource: Design Arena Leaderboards data · Nov 2026
MCP Atlas2 results
Reported by Moonshot AI · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
Kimi K2.7 Code official table · Kimi Code CLI · thinking + preserve_thinking · temp 1.0 · top_p .95 · 262,144 context · 100 tool-call budget · 32K max tokens/step · avg@3
Source: LLM Stats data · Reported by the model's developer. The source relaying it did not verify it. claim
Setting not stated
MCP-MarkSource: LLM Stats data · Reported by the model's developer. The source relaying it did not verify it. claim
MCPMark VerifiedSource: MCPMark Verified leaderboard docs
Source: OpenRouter benchmark exports
12 Jun 2026 version · 90-day results
Across 17 providers lowest 59.3% median 72.3% highest 73.3%OpenRouter's automatic routing 73.7%
Show results from 17 providers and OpenRouter's automatic routingHide results from 17 providers and OpenRouter's automatic routing
  • OpenRouter's automatic routing
  • Phala
  • Moonshot AI
  • Nebius
  • Inceptron
  • Ambient
  • Fireworks
  • StreamLake
  • Cloudflare
  • Alibaba
  • Venice
  • SiliconFlow
  • Novita
  • ModelRun
  • CoreWeave
  • Tenstorrent
  • GMICloud
  • BaseTen

Coding

Arena ELO WebdevSource: LMArena data · Nov 2026
Kimi Code Bench 2.0Kimi K2.7 Code official table · Kimi Code CLI · thinking + preserve_thinking · temp 1.0 · top_p .95 · 262,144 context · Reported by Moonshot AI · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
Reported by Moonshot AI · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
Kimi K2.7 Code official table · Kimi Code CLI · thinking + preserve_thinking · temp 1.0 · top_p .95 · 262,144 context · official 30-task subset · five-hour agent budget
Source: LLM Stats data · Reported by the model's developer. The source relaying it did not verify it. claim
Setting not stated
ValsSource: Vals AI

General

Science

GPQA Diamond20 results
Source: Epoch AI Benchmarking Hub data
Setting not stated
Source: OpenRouter benchmark exports
12 Jun 2026 version · 90-day results
Across 18 providers lowest 72.6% median 88.3% highest 92.4%OpenRouter's automatic routing 87.1%
Show results from 18 providers and OpenRouter's automatic routingHide results from 18 providers and OpenRouter's automatic routing
  • Venice
  • ModelRun
  • Fireworks
  • Cloudflare
  • Novita
  • Moonshot AI
  • Nebius
  • Inceptron
  • CoreWeave
  • Phala
  • SiliconFlow
  • Ambient
  • GMICloud
  • OpenRouter's automatic routing
  • StreamLake
  • BaseTen
  • Alibaba
  • DeepInfra
  • Tenstorrent

Human preference

Design Arena EloSource: Design Arena Leaderboards data · Nov 2026

Math

Reasoning

Factuality

Speech

Other

Reported by Moonshot AI · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
Kimi K2.7 Code official table · Kimi Code CLI · thinking + preserve_thinking · temp 1.0 · top_p .95 · 262,144 context · OpenClaw · 17 scenarios · 610 checkpoints · avg@3
Source: LLM Stats data · Reported by the model's developer. The source relaying it did not verify it. claim
Setting not stated
MCPMark-VerifiedKimi K2.7 Code official table · Kimi Code CLI · thinking + preserve_thinking · temp 1.0 · top_p .95 · 262,144 context · 100-step tool-call budget · 32K max tokens/step · avg@3 · Reported by Moonshot AI · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
ProgramBench · Metric UnspecifiedKimi K2.7 Code official table · Kimi Code CLI · thinking + preserve_thinking · temp 1.0 · top_p .95 · 262,144 context · Reported by Moonshot AI · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
SkillsBench · Vals6 related benchmarks · 6 results
With Skills · Accuracy
Reported by Vals AI · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
Vals AI · updated 2026-09-27 · metadata version 1 · 87 tasks in trace charts · 3 trials/task/condition · per-task mean then task average · OpenHands-CLI@ · With Skills · system=kimi/kimi-k2.7-code · provider=Moonshot AI · top_p=0.95 · max_output_tokens=32768
With Skills · Mean cost / test
Reported by Vals AI · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
Vals AI · updated 2026-09-27 · metadata version 1 · 87 tasks in trace charts · 3 trials/task/condition · per-task mean then task average · OpenHands-CLI@ · With Skills · system=kimi/kimi-k2.7-code · provider=Moonshot AI · top_p=0.95 · max_output_tokens=32768
With Skills · Mean duration / test
Reported by Vals AI · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
Vals AI · updated 2026-09-27 · metadata version 1 · 87 tasks in trace charts · 3 trials/task/condition · per-task mean then task average · OpenHands-CLI@ · With Skills · system=kimi/kimi-k2.7-code · provider=Moonshot AI · top_p=0.95 · max_output_tokens=32768
Without Skills · Accuracy
Reported by Vals AI · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
Vals AI · updated 2026-09-27 · metadata version 1 · 87 tasks in trace charts · 3 trials/task/condition · per-task mean then task average · OpenHands-CLI@ · Without Skills · system=kimi/kimi-k2.7-code · provider=Moonshot AI · top_p=0.95 · max_output_tokens=32768
Without Skills · Mean cost / test
Reported by Vals AI · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
Vals AI · updated 2026-09-27 · metadata version 1 · 87 tasks in trace charts · 3 trials/task/condition · per-task mean then task average · OpenHands-CLI@ · Without Skills · system=kimi/kimi-k2.7-code · provider=Moonshot AI · top_p=0.95 · max_output_tokens=32768
Without Skills · Mean duration / test
Reported by Vals AI · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
Vals AI · updated 2026-09-27 · metadata version 1 · 87 tasks in trace charts · 3 trials/task/condition · per-task mean then task average · OpenHands-CLI@ · Without Skills · system=kimi/kimi-k2.7-code · provider=Moonshot AI · top_p=0.95 · max_output_tokens=32768
WildClawBench6 related benchmarks · 6 results
WildClawBench
Source: Hugging Face Official Benchmarks data · Reported by the model's developer. The source relaying it did not verify it. claim
Multimodal (26 tasks)
Reported by WildClawBench / InternLM · via BenchBoard docs
OpenClaw · multimodal-input split · 26 tasks
Overall
Reported by WildClawBench / InternLM · via BenchBoard docs
OpenClaw · full 60-task suite
Pure-text (34 tasks)
Reported by WildClawBench / InternLM · via BenchBoard docs
OpenClaw · pure-text split · 34 tasks
Total Cost
Reported by WildClawBench / InternLM · via BenchBoard docs
OpenClaw · full 60-task suite
Total Elapsed
Reported by WildClawBench / InternLM · via BenchBoard docs
OpenClaw · full 60-task suite

Data credits: Epoch AI Benchmarking Hub, CC BY 4.0 (licence) · LMArena, CC BY 4.0 (licence) · Data by LLM Stats · OpenRouter public benchmark leaderboard and search-lane exports, CC BY 4.0. Fractional source scores converted to percentages; immutable dated snapshots. (licence) · BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links. · Hugging Face official benchmark leaderboards; licence varies by benchmark · MCPMark (eval-sys), mcpmark.ai Verified leaderboard. Single-run Pass@1 fractions shown as percentages.

Around the web

  • Kimi K2 - GroqDocs
    Built on a Mixture-of-Experts (MoE) architecture with 1 trillion total parameters and 32 billion activated …Built on a Mixture-of-Experts (MoE) architecture with 1 trillion total parameters and 32 billion activated parameters. Features 384 experts with 8 experts selected per token, optimized for efficient …

    console.groq.com

    Docs
  • GitHub - MoonshotAI/Kimi-K2: Kimi K2 is the large language model series developed by Moonshot AI team · GitHub
    ## 1. Model Introduction Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model with 32 …## 1. Model Introduction Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model with 32 billion activated parameters and 1 trillion total parameters. Trained with the Muon optimizer, K…

    github.com

    Repository
  • moonshotai/Kimi-K2-Instruct · Hugging Face
    Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model with 32 billion activated parameters and …Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model with 32 billion activated parameters and 1 trillion total parameters. Trained with the Muon optimizer, Kimi K2 achieves exception…

    huggingface.co

    Model card
  • [2507.20534] Kimi K2: Open Agentic Intelligence
    > Abstract:We introduce Kimi K2, a Mixture-of-Experts (MoE) large language model with 32 billion activated …> Abstract:We introduce Kimi K2, a Mixture-of-Experts (MoE) large language model with 32 billion activated parameters and 1 trillion total parameters. We propose the MuonClip optimizer, which improve…

    arxiv.org

    Paper
  • moonshotai/Kimi-K2.6 · Hugging Face
    How to use moonshotai/Kimi-K2.6 with Docker Model Runner: Kimi K2.6 Chat Homepage Hugging Face Twitter …How to use moonshotai/Kimi-K2.6 with Docker Model Runner: Kimi K2.6 Chat Homepage Hugging Face Twitter Follow Discord ModelScope) License 🤗 huggingchat · 📰 Tech Blog ## 1. Model Introduction Kimi K…

    huggingface.co

    Model card
  • Kimi K2 Thinking - API Pricing & Benchmarks | OpenRouter
    Kimi K2 Instruct is a large-scale Mixture-of-Experts (MoE) language model developed by Moonshot AI, featuring …Kimi K2 Instruct is a large-scale Mixture-of-Experts (MoE) language model developed by Moonshot AI, featuring 1 trillion total parameters with 32 billion active per forward pass. It is optimized for …

    openrouter.ai

  • Kimi K2: Open Agentic Intelligence
    it is within your reach. Today, we are open-sourcing: - Kimi-K2-Base: The foundation model, a …it is within your reach. Today, we are open-sourcing: - Kimi-K2-Base: The foundation model, a strong start for researchers and builders who want full control for fine-tuning and custom solutions. - K…

    moonshotai.github.io

  • Kimi (AI) - Wikipedia
    In July 2025, Moonshot AI released Kimi K2, a 1 trillion parameter model with 32 …In July 2025, Moonshot AI released Kimi K2, a 1 trillion parameter model with 32 billion active parameters which was open sourced under a modified MIT license.( On 9 September 2025, Moonshot AI relea…

    en.wikipedia.org

  • Kimi K2: A Guide With 6 Practical Examples | DataCamp
    Kimi K2 is a language model designed by Moonshot AI, featuring 32 billion activated parameters …Kimi K2 is a language model designed by Moonshot AI, featuring 32 billion activated parameters within an extensive framework of 1 trillion total parameters. Kimi K2 is optimized for agentic tasks. It…

    datacamp.com

  • Alibaba-backed Moonshot releases Kimi K2 AI rivaling ChatGPT, Claude
    In its release announcement on social media platforms X and GitHub, Moonshot claimed Kimi K2 …In its release announcement on social media platforms X and GitHub, Moonshot claimed Kimi K2 surpassed Claude Opus 4 on two benchmarks, and had better overall performance than OpenAI's coding-focused…

    cnbc.com

Also known as: @cf/moonshotai/kimi-k2.7-code, aihubmix/kimi-k2.7-code-highspeed, alibaba/kimi-k2.7-code, azure_ai/fw-kimi-k2.7-code, azure_ai/kimi-k2.7-code, baseten/moonshotai/kimi-k2.7-code, cline-pass/kimi-k2.7-code, cloudflare/@cf/moonshotai/kimi-k2.7-code, dashscope/kimi-k2.7-code, databricks-kimi-k2-7-code, deepinfra/moonshotai/kimi-k2.7-code, empiriolabs/kimi-k2-7-code, empiriolabs/kimi-k2-7-code-highspeed, fireworks_ai/accounts/fireworks/models/kimi-k2p7-code, fireworks_ai/accounts/fireworks/routers/kimi-k2p7-code-fast, fireworks_ai/kimi-k2p7-code, fireworks_ai/kimi-k2p7-code-fast, gmi/moonshotai/kimi-k2.7-code, gmi/moonshotai/kimi-k2.7-code-highspeed, inceptron/kimi-k2.7-code, kimi-k2-7-code, kimi-k2-7-code-highspeed, kimi-k2.7-code-highspeed, kimi-k2.7-code-nitro, kimi-k2.7-code:web, kimi-k2.7-code@eu, kimi-k2p7-code, kimi/kimi-k2.7-code, moonshot-ai/kimi-k2.7-code, moonshot/kimi-k2-7-code, moonshot/kimi-k2-7-code-highspeed, moonshot/kimi-k2.7-code, moonshot/kimi-k2.7-code-highspeed, moonshotai/kimi-k2.7-code, moonshotai/kimi-k2.7-code-highspeed, nebius/moonshotai/kimi-k2.7-code, novita/kimi-k2.7-code, novita/moonshotai/kimi-k2.7-code, opencode_go/kimi-k2.7-code, perplexity/kimi-k2.7-code, qwen_ai_platform/kimi-k2.7-code, qwencloud/kimi-k2.7-code, scx-ai-gp/kimi-k2.7-code, streamlake/kimi-k2.7-code, tencent/kimi-k2.7-code, tencent/kimi-k2.7-code-highspeed, tensorx/kimi-k2.7-code, together_ai/moonshotai/kimi-k2.7-code, wandb/moonshotai/kimi-k2.7-code

Like this one. Tap or hover any dotted value, label or icon to see where it comes from and how it is measured. open details when tapped or hovered. How this data works