{"meta":{"title":"AI 모델 비교","intro":"사용 가능한 AI 모델을 Copilot Chat 비교하고 작업에 가장 적합한 모델을 선택합니다.","product":"GitHub Copilot","breadcrumbs":[{"href":"/ko/copilot","title":"GitHub Copilot"},{"href":"/ko/copilot/reference","title":"참조"},{"href":"/ko/copilot/reference/ai-models","title":"AI 모델"},{"href":"/ko/copilot/reference/ai-models/model-comparison","title":"모델 비교"}],"documentType":"article"},"body":"# AI 모델 비교\n\n사용 가능한 AI 모델을 Copilot Chat 비교하고 작업에 가장 적합한 모델을 선택합니다.\n\n## 에 대한 AI 모델 비교 GitHub Copilot\n\nGitHub Copilot 는 다양한 기능을 갖춘 여러 AI 모델을 지원합니다. 선택한 모델은 인라인 제안별 응답 Copilot Chat 의 품질 및 Copilot 관련성에 영향을 줍니다. 일부 모델은 대기 시간이 더 짧고, 또 다른 모델은 환각이 적거나 또는 특정 작업에서 성능 향상을 제공합니다. 이 가이드는 모델 이름뿐만 아니라 작업에 따라 가장 적합한 모델을 선택하는 데 도움이 됩니다.\n\n> \\[!NOTE]\n>\n> * 서로 다른 모델은 토큰 가격 책정에 따라 서로 다른 속도로 사용합니다 AI credits . 세부 정보는 [GitHub Copilot에 대한 모델 및 가격 책정](/ko/copilot/reference/copilot-billing/models-and-pricing)(을)를 참조하세요.\n> * 모델을 선택할 때 **자동** 옵션을 선택하면 모델 가용성 및 작업 복잡성에 따라 작업에 대한 최적의 모델이 자동으로 선택됩니다.  유료 Copilot 요금제인 경우 **자동** 옵션을 사용할 때 모델 비용 할인을 받을 수 있습니다.\n>   [정보 Copilotauto model selection](/ko/copilot/concepts/models/auto-model-selection)을(를) 참조하세요.\n\n### 작업별 권장 모델\n\n이 표를 사용하면 적합한 모델을 빠르게 찾을 수 있습니다. 아래 섹션에서 자세한 내용을 참조하세요.\n\n| 모델         | 작업 영역 | 주 용도에서 뛰어남 | 추가 참고 자료 |\n| ---------- | ----- | ---------- | -------- |\n|            |       |            |          |\n| GPT-5 mini |       |            |          |\n\n```\n                               | General-purpose coding and writing | Fast, accurate code completions and explanations        | [GPT-5 mini model card](https://cdn.openai.com/gpt-5-system-card.pdf) |\n```\n\n\\|  |\n\\| GPT-5.3-Codex\n\\| Agentic software development | Agentic tasks        | [GPT-5.3-Codex model card](https://deploymentsafety.openai.com/gpt-5-3-codex) |\n\\|  |\n\\| GPT-5.4\n\\| Deep reasoning and debugging | Multi-step problem solving and architecture-level code analysis        | [GPT-5.4 model card](https://deploymentsafety.openai.com/gpt-5-4-thinking/introduction) |\n\\|  |\n\\| GPT-5.4 mini\n\\| Agentic software development | Codebase exploration and is especially effective when using grep-style tools        | [GPT-5.4 mini model card](https://deploymentsafety.openai.com/gpt-5-4-thinking/gpt-5-4-thinking.pdf) |\n\\|  |\n\\| GPT-5.5\n\\| Deep reasoning and debugging | Multi-step problem solving and architecture-level code analysis        | [GPT-5.5 model card](https://deploymentsafety.openai.com/gpt-5-5/gpt-5-5.pdf) |\n\\|  |\n\\| GPT-5.6 Luna\n\\| Fast help with simple or repetitive tasks | Quick, cost-efficient responses for smaller, faster coding tasks        | [GPT-5.6 model card](https://deploymentsafety.openai.com/gpt-5-6/gpt-5-6.pdf) |\n\\|  |\n\\| GPT-5.6 Sol\n\\| Deep reasoning and debugging | Complex reasoning over large codebases and long-running agentic work        | [GPT-5.6 model card](https://deploymentsafety.openai.com/gpt-5-6/gpt-5-6.pdf) |\n\\|  |\n\\| GPT-5.6 Terra\n\\| General-purpose coding and agent tasks | Balanced everyday interactive and agentic coding        | [GPT-5.6 model card](https://deploymentsafety.openai.com/gpt-5-6/gpt-5-6.pdf) |\n\\|  |\n\\| Claude Fable 5\n\\| Long-horizon, autonomous coding and knowledge-work | First-attempt correctness through upfront reasoning, aggressive parallel tool batching, and proactive verification of pre-existing test state before debugging        | [Claude Fable 5 model card](https://anthropic.com/claude-fable-5-mythos-5-system-card) |\n\\|  |\n\\| Claude Fable 5.1\n\\| Long-horizon, autonomous coding and knowledge-work | Substantial, long-running coding tasks, including deep codebase research, feature development, and complex agentic workflows        | [Claude Fable 5.1 model card](https://www-cdn.anthropic.com/0339e6a7c5c7b87f5c07798616dc32c215d14235/Claude%20Fable%205.1%20&%20Claude%20Mythos%205.1%20System%20Card.pdf) |\n\\|  |\n\\| Claude Haiku 4.5\n\\| Fast help with simple or repetitive tasks | Fast, reliable answers to lightweight coding questions        | [Claude Haiku 4.5 model card](https://assets.anthropic.com/m/99128ddd009bdcb/Claude-Haiku-4-5-System-Card.pdf) |\n\\|  |\n\\| Claude Opus 4.7\n\\| Deep reasoning and debugging | Complex problem-solving challenges, sophisticated reasoning        | [Claude Opus 4.7 model card](https://cdn.sanity.io/files/4zrzovbb/website/037f06850df7fbe871e206dad004c3db5fd50340.pdf) |\n\\|  |\n\\| Claude Opus 4.8\n\\| Deep reasoning and debugging | Complex problem-solving challenges, sophisticated reasoning        | [Claude Opus 4.8 model card](https://www-cdn.anthropic.com/0b4915911bb0d19eca5b5ee635c80fef830a37ea.pdf) |\n\\|  |\n\\| Claude Opus 4.8 (fast mode) (preview)\n\\| Deep reasoning and debugging | Complex problem-solving challenges, sophisticated reasoning        | [Claude Opus 4.8 model card](https://www-cdn.anthropic.com/0b4915911bb0d19eca5b5ee635c80fef830a37ea.pdf) |\n\\|  |\n\\| Claude Opus 5\n\\| Deep reasoning and debugging | Complex problem-solving challenges, sophisticated reasoning        | [Claude Opus 5 model card](https://www-cdn.anthropic.com/c5fbac3f0b1280a933ebd26d3cb8bb9f5bdeaf48/Claude%20Opus%205%20System%20Card.pdf) |\n\\|  |\n\\| Claude Sonnet 4.5\n\\| General-purpose coding and agent tasks | Complex problem-solving challenges, sophisticated reasoning        | [Claude Sonnet 4.5 model card](https://assets.anthropic.com/m/12f214efcc2f457a/original/Claude-Sonnet-4-5-System-Card.pdf) |\n\\|  |\n\\| Claude Sonnet 4.6\n\\| General-purpose coding and agent tasks | Complex problem-solving challenges, sophisticated reasoning        | [Claude Sonnet 4.6 model card](https://www-cdn.anthropic.com/78073f739564e986ff3e28522761a7a0b4484f84.pdf) |\n\\|  |\n\\| Claude Sonnet 5\n\\| General-purpose coding and agent tasks | Complex problem-solving challenges, sophisticated reasoning        | [Claude Sonnet 5 model card](https://www-cdn.anthropic.com/9e6a1044980d8c4ed85669faf9c2a8342e2e9f1e/Claude%20Sonnet%205%20System%20Card.pdf) |\n\\|  |\n\\| Gemini 3.1 Pro\n\\| Deep reasoning and debugging | Effective and efficient edit-then-test loops with high tool precision        | [Gemini 3.1 Pro model card](https://storage.googleapis.com/deepmind-media/Model-Cards/Gemini-3-1-Pro-Model-Card.pdf) |\n\\|  |\n\\| Gemini 3.5 Flash\n\\| Fast help with simple or repetitive tasks | Fast, reliable answers to lightweight coding questions        | [Gemini 3.5 Flash model card](https://storage.googleapis.com/deepmind-media/Model-Cards/Gemini-3-5-Flash-Model-Card.pdf) |\n\\|  |\n\\| Gemini 3.6 Flash\n\\| Fast help with simple or repetitive tasks | Fast, reliable answers to lightweight coding questions        | [Gemini 3.6 Flash model card](https://storage.googleapis.com/deepmind-media/Model-Cards/Gemini-3-6-Flash-Model-Card.pdf) |\n\\|  |\n\\| Gemini 3.7 Flash\n\\| Fast help with simple or repetitive tasks | Fast, reliable answers to lightweight coding questions        | [Gemini 3.7 Flash model card](https://storage.googleapis.com/deepmind-media/Model-Cards/Gemini-3-7-Flash-Model-Card.pdf) |\n\\|  |\n\\| Gemini 3.8 Flash\n\\| Fast help with simple or repetitive tasks | Fast, reliable answers to lightweight coding questions        | [Gemini 3.8 Flash model card](https://storage.googleapis.com/deepmind-media/Model-Cards/Gemini-3-8-Flash-Model-Card.pdf) |\n\\|  |\n\\| MAI-Code-1-Flash\n\n[^mai-code-1-flash]                                   | General-purpose coding and writing | Fast, accurate code completions and explanations        | [MAI-Code-1-Flash model card](https://aka.ms/mai-code-1-flash-modelcard) |\n\\|  |\n\\| MAI-Code-1.1-Flash\n\n[^mai-code-1-flash]                                   | General-purpose coding and writing, image understanding | Fast code completions and explanations, instruction following, tool use        | [MAI-Code-1.1-Flash model card](https://microsoft.ai/pdf/MAI-Code-1.1-Flash-Model-Card.PDF) |\n\\|  |\n\\| Grok 4.5\n\\| General-purpose coding and agent tasks | Complex problem-solving challenges, sophisticated reasoning        | [Grok 4.5 model card](https://media.x.ai/v1/website/card-7f81d41b.pdf) |\n\\|  |\n\\| Grok 4.6\n\\| General-purpose coding and agent tasks | Complex problem-solving challenges, sophisticated reasoning        | [Grok 4.6 model card](https://media.x.ai/v1/website/card-4p6-4cd2dc57.pdf) |\n\\|  |\n\\| Qwen2.5\n\\| General-purpose coding and writing | Code generation, reasoning, and code repair / debugging        | [Qwen2.5 model card](https://arxiv.org/pdf/2409.12186) |\n\\|  |\n\\| Raptor mini\n\\| General-purpose coding and writing | Fast, accurate code completions and explanations        | Coming soon |\n\\|  |\n\\| Kimi K2.7 Code\n\\| General-purpose coding and agent tasks | Fast, reliable answers to lightweight coding questions        | [Kimi K2.7 Code model card](https://huggingface.co/moonshotai/Kimi-K2.7-Code) |\n\\|  |\n\\| Kimi K3\n[^kimi-k3]                                   | Agentic coding and long-context work | Multi-step agent tasks across large codebases        | [Kimi K3 model card](https://huggingface.co/moonshotai/Kimi-K3) |\n\\|                                        |\n\n## 작업: 범용 코딩 및 쓰기\n\n품질, 속도, 비용 효율성의 균형이 필요한 일반적인 개발 작업에 이러한 모델을 사용합니다. 이러한 모델은 특별한 요구 사항이 없는 경우 기본값으로 사용하기 좋은 모델입니다.\n\n| 모델               | 적합한 이유                                                                      |\n| ---------------- | --------------------------------------------------------------------------- |\n| GPT-5.3-Codex    | 긴 지침 없이 기능, 테스트, 디버깅, 리팩터, 검토와 같은 복잡한 엔지니어링 작업에 대한 고품질 코드를 제공합니다.           |\n| GPT-5 mini       | 대부분의 코딩 및 쓰기 작업에 대한 신뢰할 수 있는 기본값입니다. 빠르고 정확하며 다양한 언어 및 프레임워크와 잘 호환됩니다.      |\n| GPT-5.6 Terra    | 일상적인 대화형 및 에이전트 코딩을 위한 균형 잡힌 다라운드 선택.                                       |\n| Raptor mini      | 빠르고 정확한 인라인 제안 및 설명을 위해 전문화되었습니다.                                           |\n| MAI-Code-1-Flash | 강력한 명령 팔로우 및 적응 추론을 통해 일상적인 코딩 작업, 쓰기 및 멀티 턴 개발 워크플로에 대한 신뢰할 수 있는 기본값이 됩니다. |\n\n### 이러한 모델을 사용하는 경우\n\n이러한 모델 중 하나를 사용하여 다음 작업을 수행합니다.\n\n* 함수, 짧은 파일, 또는 코드 변경(diff)을 작성하거나 검토합니다.\n* 문서, 메모, 요약 생성\n* 오류 또는 예기치 않은 동작 빠르게 설명\n* 영어가 아닌 프로그래밍 환경에서 작업\n\n### 다른 모델을 사용하는 경우\n\n복잡한 리팩터링, 아키텍처 결정, 다단계 논리 작업을 수행하는 경우, [심층 추론 및 디버깅](#task-deep-reasoning-and-debugging) 모델을 고려하세요. 반복적인 편집 또는 일회성 코드 제안과 같은 더 빠르고 간단한 작업은 [단순하거나 반복적인 작업에 대한 빠른 지원](#task-fast-help-with-simple-or-repetitive-tasks)을 참조하세요.\n\n## 작업: 단순하거나 반복적인 작업에 대한 빠른 지원\n\n이러한 모델은 속도와 응답성에 최적화되어 있습니다. 빠른 편집, 유틸리티 함수, 구문 도움말, 간단한 프로토타입 작성에 이상적입니다. 불필요한 심층적 또는 긴 추론 과정을 기다리지 않고 빠른 답변을 얻을 수 있습니다.\n\n### 권장 모델\n\n| 모델               | 적합한 이유                                                                       |\n| ---------------- | ---------------------------------------------------------------------------- |\n| GPT-5.6 Luna     | 더 작고 빠른 작업을 위한 가볍고 비용 효율적인 옵션입니다. GPT-5.6 제품군의 최저 비용 모델입니다.                  |\n| Claude Haiku 4.5 | 빠른 응답과 고품질 출력의 균형을 유지합니다. 작은 작업 및 간단한 코드 설명에 이상적입니다.                         |\n| MAI-Code-1-Flash | 적응 효율성으로 빠른 코딩 작업을 처리하고, 간단한 요청에 대해 간결하게 유지하며, 불필요한 깊이 없이 빠르고 정확한 응답을 제공합니다. |\n\n### 이러한 모델을 사용하는 경우\n\n이러한 모델 중 하나를 사용하여 다음 작업을 수행합니다.\n\n* 작은 함수 또는 유틸리티 코드를 작성하거나 편집합니다.\n* 구문 또는 언어에 대한 간단한 질문을 하세요.\n* 최소한의 설정으로 아이디어를 프로토타입화합니다.\n* 간단한 프롬프트 또는 편집에 대한 빠른 피드백을 받습니다.\n\n### 다른 모델을 사용하는 경우\n\n복잡한 리팩터링, 아키텍처 결정, 다단계 논리 작업을 수행하는 경우, [심층 추론 및 디버깅](#task-deep-reasoning-and-debugging)을 참조하세요.\n더 강력한 범용 추론 또는 보다 구조화된 출력이 필요한 작업은 [범용 코딩 및 쓰기](#task-general-purpose-coding-and-writing)를 참조하세요.\n\n## 작업: 심층 추론 및 디버깅\n\n이러한 모델은 단계별 추론, 복잡한 의사 결정, 높은 컨텍스트 인식이 필요한 작업을 위해 설계되었습니다. 구조적 분석, 신중한 코드 생성, 다중 파일 이해가 필요할 때 효과적입니다.\n\n### 권장 모델\n\n| 모델                         | 적합한 이유                                                                          |\n| -------------------------- | ------------------------------------------------------------------------------- |\n| GPT-5 mini                 | GPT-5보다 더 빠른 응답과 낮은 리소스 사용량으로 깊은 추론 및 디버깅을 제공합니다. 대화형 세션 및 단계별 코드 분석에 적합합니다.    |\n| GPT-5.5                    | 복잡한 추론, 코드 분석, 기술 의사 결정에 적합합니다.                                                 |\n| GPT-5.6 Sol                | GPT-5.6 제품군에서 가장 높은 추론 한도입니다. 큰 코드베이스에 대한 복잡한 추론과 까다로운 장기 실행 에이전트 작업에 가장 적합합니다. |\n| Claude Sonnet 4.6          | Sonnet 4.5의 구현을 보다 안정적으로 하고, 압박 속에서도 더 스마트한 추론을 통해 개선합니다.                       |\n| Claude Opus 4.7            | Anthropic의 가장 강력한 모델입니다.                                                        |\n| Claude Opus 4.6을(를) 개선합니다. |                                                                                 |\n| Gemini 3.1 Pro             | 긴 컨텍스트에 대한 고급 추론과 과학적 기술적 분석에 뛰어납니다.                                            |\n\n### 이러한 모델을 사용하는 경우\n\n이러한 모델 중 하나를 사용하여 다음 작업을 수행합니다.\n\n* 여러 파일에서 컨텍스트와 관련된 복잡한 문제 디버그\n* 대규모 또는 상호 연결된 코드베이스 리팩터링\n* 여러 계층에 걸친 기능이나 아키텍처 계획\n* 라이브러리, 패턴, 워크플로 간의 장단점 비교 및 조정\n* 로그, 성능 데이터, 시스템 동작 분석\n\n### 다른 모델을 사용하는 경우\n\n빠른 반복 또는 간단한 작업은 [단순하거나 반복적인 작업에 대한 빠른 지원](#task-fast-help-with-simple-or-repetitive-tasks)을 참조하세요.\n일반적인 개발 워크플로 또는 콘텐츠 생성은 [범용 코딩 및 쓰기](#task-general-purpose-coding-and-writing)를 참조하세요.\n\n## 작업: 시각적 자료(다이어그램, 스크린샷) 다루기\n\n스크린샷, 다이어그램, UI 구성 요소 또는 기타 시각적 입력에 대한 질문을 할 때 이러한 모델을 사용합니다. 이러한 모델은 다중 모드 입력을 지원하며 프런트 엔드 작업 또는 시각적 디버깅에 적합합니다.\n\n| 모델                | 적합한 이유                                                                                             |\n| ----------------- | -------------------------------------------------------------------------------------------------- |\n| GPT-5 mini        | 대부분의 코딩 및 쓰기 작업에 대한 신뢰할 수 있는 기본값입니다. 빠르고 정확하며 시각적 추론 작업을 위한 다중 모달 입력을 지원합니다. 언어 및 프레임워크에서 잘 작동합니다. |\n| Claude Sonnet 4.6 | Sonnet 4.5의 구현을 보다 안정적으로 하고, 압박 속에서도 더 스마트한 추론을 통해 개선합니다.                                          |\n| Gemini 3.1 Pro    | 복잡한 코드 생성, 디버깅, 연구 워크플로에 적합한 심층 추론 및 디버깅이 가능합니다.                                                   |\n\n### 이러한 모델을 사용하는 경우\n\n이러한 모델 중 하나를 사용하여 다음 작업을 수행합니다.\n\n* 다이어그램, 스크린샷, UI 구성 요소에 대해 질문합니다.\n* 시각적 초안 또는 워크플로에 대한 피드백을 받습니다.\n* 시각적 컨텍스트에서 프런트 엔드 동작을 이해합니다.\n\n> \\[!TIP]\n> 이미지 입력(예: 코드 편집기)을 지원하지 않는 컨텍스트에서 모델을 사용하는 경우 시각적 추론 이점이 표시되지 않습니다. MCP 서버를 사용하면 시각적 입력에 간접적으로 액세스할 수 있습니다.\n> [MCP(모델 컨텍스트 프로토콜) 서버를 사용하여 GitHub Copilot Chat 확장](/ko/copilot/how-tos/provide-context/use-mcp-in-your-ide/extend-copilot-chat-with-mcp)을(를) 참조하세요.\n\n### 다른 모델을 사용하는 경우\n\n작업에 심층 추론 또는 대규모 리팩터링이 포함된 경우 [심층 추론 및 디버깅](#task-deep-reasoning-and-debugging)의 모델을 고려합니다. 텍스트 전용 작업이나 간단한 코드 편집의 경우 [단순하거나 반복적인 작업에 대한 빠른 지원](#task-fast-help-with-simple-or-repetitive-tasks)을 참조하세요.\n\n## 모델별 고려 사항\n\n일부 모델에는 작업에 대해 선택하기 전에 이해하는 데 유용한 동작, 제한 사항 또는 보호 장치가 있습니다.\n\n### Kimi K3\n\nKimi K3 는 긴 컨텍스트, 다단계 코딩 및 에이전트 워크플로를 위해 설계되었습니다. 시험판 테스트에서 모델은 특정 고위험 프롬프트에 높은 위험을 보였고 중요한 토픽과 관련된 요청을 거부하는 다른 모델보다 덜 일관되었습니다. 이러한 동작은 학습 후 모델의 안전성 및 맞춤의 차이를 반영할 수 있습니다. 식별된 위험을 완화하는 데 도움이 되는 추가 안전 장치를 GitHub Copilot 배포했습니다. 모든 모델과 마찬가지로 기업은 특정 사용 사례 및 요구 사항에 비추어 모델 기능, 제한 사항 및 안전 장치를 평가해야 합니다.\n\n## 다음 단계\n\n[^mai-code-1-flash]: MAI models are continuously improving models. Performance and behavior may evolve over time as new checkpoints are released.\n\n[^kimi-k3]: For important information about Kimi K3 behavior and safeguards, see [Model-specific considerations](#kimi-k3).\n\n적절한 모델을 선택하면 Copilot를 최대한 활용할 수 있습니다. 어떤 모델을 사용할지 잘 모르는 경우 다음과 같은 GPT-5 mini범용 옵션으로 시작한 다음 필요에 따라 조정합니다.\n\n* 자세한 모델 사양 및 가격은 [GitHub Copilot에서 지원되는 AI 모델](/ko/copilot/reference/ai-models/supported-models)을(를) 참조하세요.\n* 다양한 모델을 사용하는 방법에 대한 더 많은 예를 보려면 [다른 작업을 사용하여 AI 모델 비교](/ko/copilot/tutorials/compare-ai-models)을(를) 참조하세요.\n* 모델 간 서로 전환하려면 [GitHub Copilot 채팅에 대한 AI 모델 변경](/ko/copilot/how-tos/use-ai-models/change-the-chat-model) 또는 [GitHub Copilot 인라인 제안을 위한 AI 모델 변경](/ko/copilot/how-tos/use-ai-models/change-the-completion-model)을(를) 참조하세요.\n* 다양한 AI 모델을 제공하는 방법을 Copilot Chat 알아보려면 [GitHub Copilot 위한 모델 호스팅](/ko/copilot/reference/ai-models/model-hosting)을 참조하세요."}