test

Okay, the user is asking for a list of open-source LLMs in Korean. Let me recall the main ones.

First, there’s Llama series by Meta. Llama 3 is the latest, but they might not have released the full details yet. Wait, Llama 2 is available, and Llama 3 might be upcoming. I should mention both but clarify the status.

Then there’s Mistral AI’s models, like Mistral 7B. They have some open-source models. Wait, Mistral has a few, like Mixtral 8x7B, which is a Mixture of Experts model. But I need to check if they’re fully open-source or have some restrictions.

Qwen by Tongyi Lab is another one. Qwen is open-source, and they have different versions like Qwen-Max, but maybe the open-source ones are Qwen or Qwen2. Wait, Qwen is the main one they released.

Then there’s Falcon, developed by TII. Falcon-40B and Falcon-180B are available. But I think Falcon is open-source under Apache 2.0.

Bloom by BigScience is another one. It’s a large multilingual model. I should mention that.

Also, maybe the Llama 2 models are available through Hugging Face. Need to confirm if they’re open-source. Yes, Meta released Llama 2 under a license that allows commercial use, but requires a request.

Oh, and there’s the MPT series by MosaicML. MPT-7B and others. They are open-source and available on Hugging Face.

Then there’s the Chinese models like Qwen, but also maybe Baichuan or others. Wait, Baichuan is by Baichuan Intelligent, but are they open-source? I think some of their models are, but I’m not 100% sure. Maybe better to stick to confirmed ones.

Also, the StarChat models by Hugging Face, but I’m not sure if they’re open-source. Wait, StarChat might be a different model. Maybe not.

Wait, there’s also the Yi series by 01.AI. Yi-34B is a model they released, and it’s open-source.

Also, the RWKV series, but those are more for RNNs, not LLMs. Maybe not relevant here.

Let me list the main ones:

1. Llama 2 (Meta)
2. Llama 3 (Meta, upcoming)
3. Mistral 7B (Mistral AI)
4. Mixtral 8x7B (Mistral)
5. Falcon 40B/180B (TII)
6. Bloom (BigScience)
7. Qwen (Tongyi Lab)
8. MPT-7B (MosaicML)
9. Yi-34B (01.AI)

Wait, I should check if all these are indeed open-source. For example, Mistral 7B is open-source, but Mixtral might have some restrictions. Wait, Mixtral is open-source under Apache 2.0, I think.

Also, the user is asking in Korean, so I should present the list with names and maybe a brief note on each. Need to make sure the names are correctly translated or kept in English as they’re proper nouns.

Wait, the user said “오픈소스 LLM 종류 알려줘”, so they want the types or list of open-source LLMs. So I should list them with their names and maybe a short description.

Also, check if some models are only partially open-source or have usage restrictions. For example, Llama 2 requires approval, but it’s still considered open-source.

So, compiling the list with the most prominent ones. Let me verify a few:

• Llama 2: Yes, open-source, Meta.
• Mistral 7B: Mistral AI, open-source.
• Falcon: Yes, Apache 2.0.
• Bloom: Open-source.
• Qwen: Tongyi Lab, open-source.
• MPT: Open-source.
• Yi-34B: 01.AI, open-source.

Also, maybe mention Hugging Face models, but they might not be LLMs themselves. Wait, Hugging Face hosts many models, but the LLMs are from the companies above.

So, the list should include the main ones. Let me structure the answer in Korean, listing each model with a brief note.

Wait, the user wants the answer in Korean. So I need to write the names of the models as they are (since they’re proper nouns), but describe them in Korean.

For example:

• Llama 2: 메타에서 개발한 오픈소스 LLM
• Mistral 7B: 미스트랄 AI에서 출시한 오픈소스 모델

Make sure to clarify which ones are fully open-source and maybe note any restrictions (like Llama 2 requiring a license request).

Also, check if Llama 3 is released yet. As of 2023, Llama 3 might not be out, so maybe mention Llama 2 as the current one.

Another model: The Chinese models like BLOOM, but maybe not. Wait, Bloom is multilingual, but not specifically Chinese.

Wait, there’s also the InternLM series by Shanghai AI Lab, but I’m not sure if they’re open-source. InternLM-7B might be, but I need to confirm. Maybe better to stick to the ones I’m sure about.

So, final list:

1. Llama 2 (Meta)
2. Mistral 7B (Mistral AI)
3. Mixtral 8x7B (Mistral AI)
4. Falcon 40B/180B (TII)
5. Bloom (BigScience)
6. Qwen (Tongyi Lab)
7. MPT-7B (MosaicML)
8. Yi-34B (01.AI)

Wait, MPT-7B is part of the MPT series. Also, there’s MPT-7B, MPT-30B, etc.

Also, maybe mention that some models are available on Hugging Face, but that’s common for most open-source models.

Now, write the answer in Korean, listing these models with brief descriptions. Make sure to say “오픈소스” for each, and note any important points like licensing.

Check if any of these are not actually open-source. For example, Mistral 7B: Mistral AI has some models under Apache 2.0, so yes. Mixtral is also Apache 2.0.

Falcon is Apache 2.0.

Bloom is Apache 2.0.

Qwen is open-source under Apache 2.0.

MPT is Apache 2.0.

Yi-34B is under the Apache 2.0 license.

Okay, so all are open-source with Apache 2.0 or similar.

Llama 2 requires a license from Meta, but it’s still considered open-source.

So, in the answer, mention that some require approval (like Llama 2), but others are freely available.

Wait, the user just wants the types, so maybe list them without getting into the license details unless necessary.

But it’s good to note that Llama 2 requires a request, while others are directly downloadable.

So, structure the answer as a list with each model, their developer, and a short note.

Make sure to use Korean for the descriptions, but keep the model names in English.

Let me write the answer now.

다음은 주요 오픈소스 LLM(대형 언어 모델) 종류입니다. 일부 모델은 사용 승인을 필요로 할 수 있으나, 대부분 Apache 2.0 등 오픈소스 라이선스 하에 공개되었습니다.

🔹 메타 (Meta)

• Llama 2:
  메타에서 개발한 대형 모델로, Llama 2-7B, 13B, 70B 버전이 공개되었습니다.
  ⚠️ 주의: 상업적 사용을 위해서는 메타의 승인을 받아야 합니다.

• Llama 3 (예정):
  Llama 2 이후 출시 예정인 모델로, 더 강력한 성능을 목표로 합니다.

🔹 Mistral AI

• Mistral 7B:
  7B 파라미터 모델로, 효율적인 성능과 빠른 추론이 특징입니다.
• Mixtral 8x7B:
  “Mixture of Experts” 아키텍처를 사용한 모델로, 복잡한 작업에서 뛰어난 성능을 보입니다.

🔹 TII (Technology Innovation Institute)

• Falcon 40B/180B:
  40B 및 180B 파라미터 모델로, 대규모 언어 처리에 최적화되었습니다.
  📌 특징: Apache 2.0 라이선스 하에 완전히 오픈소스.

🔹 BigScience

• Bloom:
  다국어 지원(100개 이상 언어) 대형 모델로, 다양한 언어 처리 작업에 사용 가능합니다.

🔹 Tongyi Lab (알리바바 계열)

• Qwen:
  알리바바에서 개발한 중국어 및 영어 지원 모델로, Qwen, Qwen2, Qwen3 등 버전이 있습니다.
  📌 특징: Apache 2.0 라이선스로 공개되었으며, Hugging Face에서 다운로드 가능.

🔹 MosaicML

• MPT-7B/30B:
  MosaicML에서 개발한 모델로, 대규모 텍스트 생성 및 분석에 적합합니다.

🔹 01.AI

• Yi-34B:
  34B 파라미터 모델로, 중국어 및 영어 처리에 특화되었습니다.
  📌 특징: Apache 2.0 라이선스로 공개.

📌 참고 사항

• 대부분의 모델은 Hugging Face에서 직접 다운로드 가능합니다.
• Llama 2는 사용 승인 절차가 필요하므로, 공식 웹사이트에서 신청해야 합니다.
• 중국어 지원 모델은 Qwen, Yi-34B 등이 주로 사용됩니다.

이 목록은 2023년 기준으로 정리되었습니다. 추가로 궁금한 점이 있으면 언제든지 물어보세요! 😊 G