지원 모델

지원 모델 (Supported Models)

이 섹션은 Ascend NPU에서 지원되는 모델을 설명해요. 대규모 언어 모델(LLM), 멀티모달 언어 모델, Diffusion 언어 모델, 임베딩 모델, 보상 모델, 재랭크 모델을 포함해요. 주류 DeepSeek/Qwen/GLM 시리즈가 포함돼요. 비즈니스 요구에 따라 다양한 모델을 활성화하는 것을 환영해요.

출처: 문서

본문

대규모 언어 모델 (Large Language Models)

모델 모델 패밀리 A2 지원 A3 지원
Eco-Tech/DeepSeek-V4-Pro-w4a8-mtp DeepSeek ✅ ✅
Eco-Tech/DeepSeek-V4-Flash-w8a8-mtp DeepSeek ✅ ✅
sgl-npu/DeepSeek-V3.1-w8a8 DeepSeek ✅ ✅
sgl-npu/DeepSeek-V3.2-W8A8 DeepSeek ✅ ✅
sgl-npu/DeepSeek-R1-0528-W8A8 DeepSeek ✅ ✅
sgl-npu/DeepSeek-V2-Lite-W8A8 DeepSeek ✅ ✅
Qwen/Qwen3.6-35B-A3B Qwen3.6 ✅ ✅
Eco-Tech/Qwen3.6-27B-w8a8 Qwen3.6 ✅ ✅
Eco-Tech/Qwen3.5-397B-A17B-w4a8-mtp Qwen3.5 ✅ ✅
Eco-Tech/Qwen3.5-122B-A10B-w8a8-mtp Qwen3.5 ✅ ✅
Eco-Tech/Qwen3.5-35B-A3B-w8a8-mtp Qwen3.5 ✅ ✅
Eco-Tech/Qwen3.5-27B-w8a8-mtp Qwen3.5 ✅ ✅
Qwen/Qwen3.5-9B Qwen3.5 ✅ ✅
Qwen/Qwen3.5-4B Qwen3.5 ✅ ✅
Qwen/Qwen3.5-0.8B Qwen3.5 ✅ ✅
Qwen/Qwen3-30B-A3B-Instruct-2507 Qwen3 ✅ ✅
Qwen/Qwen3-32B Qwen3 ✅ ✅
Qwen/Qwen3-0.6B Qwen3 ✅ ✅
sgl-npu/Qwen3-235B-A22B-W8A8 Qwen3 ✅ ✅
Qwen/Qwen3-Next-80B-A3B-Instruct Qwen3 ✅ ✅
Eco-Tech/Qwen3-Coder-480B-A35B-Instruct-w8a8-QuaRot Qwen3 ✅ ✅
Qwen/Qwen2.5-7B-Instruct Qwen2.5 ✅ ✅
sgl-npu/QwQ-32B-W8A8 QWQ ✅ ✅
LLM-Research/Llama-4-Scout-17B-16E-Instruct Llama ✅ ✅
AI-ModelScope/Llama-3.1-8B-Instruct Llama ✅ ✅
LLM-Research/llama-2-7b Llama ✅ ✅
LLM-Research/Llama-3.2-1B-Instruct Llama ✅ ✅
mistralai/Mistral-7B-Instruct-v0.2 Mistral ✅ ✅
google/gemma-3-4b-it Gemma ✅ ✅
microsoft/Phi-4-multimodal-instruct Phi ✅ ✅
allenai/OLMoE-1B-7B-0924 OLMoE ✅ ✅
stabilityai/stablelm-2-1_6b StableLM ✅ ✅
CohereForAI/c4ai-command-r-v01 Command-R ✅ ✅
huihui-ai/grok-2 Grok ✅ ✅
ZhipuAI/chatglm2-6b ChatGLM ✅ ✅
LGAI-EXAONE/EXAONE-3.5-7.8B-Instruct ExaONE 3 ✅ ✅
xverse/XVERSE-MoE-A36B XVERSE ✅ ✅
HuggingFaceTB/SmolLM-1.7B SmolLM ✅ ✅
Eco-Tech/GLM-5.2-w8a8 GLM-5 ✅ ✅
Eco-Tech/GLM-5.1-w4a8 GLM-5 ✅ ✅
Eco-Tech/GLM-5-w4a8 GLM-5 ✅ ✅
ZhipuAI/glm-4-9b-chat GLM-4 ✅ ✅
XiaomiMiMo/MiMo-7B-RL MiMo ✅ ✅
arcee-ai/AFM-4.5B-Base Arcee AFM-4.5B ✅ ✅
Howeee/persimmon-8b-chat Persimmon ✅ ✅
inclusionAI/Ling-lite Ling ✅ ✅
ibm-granite/granite-3.1-8b-instruct Granite ✅ ✅
AI-ModelScope/dbrx-instruct DBRX (Databricks) ✅ ✅
baichuan-inc/Baichuan2-13B-Chat Baichuan 2 (7B, 13B) ✅ ✅
PaddlePaddle/ERNIE-4.5-21B-A3B-PT ERNIE-4.5 (4.5, 4.5MoE series) ✅ ✅
OpenBMB/MiniCPM3-4B MiniCPM (v3, 4B) ✅ ✅
Eco-Tech/Kimi-K2.6-w4a8 Kimi ✅ ✅
Eco-Tech/Kimi-K2.5-w4a8 Kimi ✅ ✅
moonshotai/Kimi-K2-Thinking Kimi ✅ ✅
eigen-ai-labs/gpt-oss-120b-bf16 GPTOSS ✅ ✅
allenai/OLMo-2-1124-7B-Instruct OLMo ✅ ✅
Eco-Tech/MiniMax-M2.5-w8a8-QuaRot MiniMax-M2.5 ✅ ✅
cyankiwi/MiniMax-M2-BF16 MiniMax-M2 ✅ ✅
upstage/SOLAR-10.7B-Instruct-v1.0 Solar ✅ ✅
FLM/Tele-FLM Tele FLM (52B-1T) ✅ ✅
bigcode/starcoder2-7b StarCoder2 ✅ ✅
arcee-ai/Trinity-Mini Trinity (Nano, Mini) ✅ ✅
OrionStarAI/Orion-14B-Base Orion (14B) ✅ ✅
EleutherAI/gpt-j-6b GPT-J (6B) ✅ ✅

멀티모달 언어 모델 (Multimodal Language Models)

모델 모델 패밀리 (변형) A2 지원 A3 지원
Qwen/Qwen2.5-VL-3B-Instruct Qwen2.5-VL ✅ ✅
Qwen/Qwen2.5-VL-72B-Instruct Qwen2.5-VL ✅ ✅
Qwen/Qwen3-VL-30B-A3B-Instruct Qwen3-VL ✅ ✅
Qwen/Qwen3-VL-8B-Instruct Qwen3-VL ✅ ✅
Qwen/Qwen3-VL-4B-Instruct Qwen3-VL ✅ ✅
Qwen/Qwen3-VL-235B-A22B-Instruct Qwen3-VL ✅ ✅
deepseek-ai/deepseek-vl2 DeepSeek-VL2 ✅ ✅
deepseek-ai/Janus-Pro-1B Janus-Pro (1B, 7B) ✅ ✅
deepseek-ai/Janus-Pro-7B Janus-Pro (1B, 7B) ✅ ✅
OpenBMB/MiniCPM-V-2_6 MiniCPM-V / MiniCPM-o ✅ ✅
OpenBMB/MiniCPM-o-2_6 MiniCPM-V / MiniCPM-o ✅ ✅
google/gemma-3-4b-it Gemma 3 (Multimodal) ✅ ✅
mistralai/Mistral-Small-3.1-24B-Instruct-2503 Mistral-Small-3.1-24B ✅ ✅
microsoft/Phi-4-multimodal-instruct Phi-4-multimodal-instruct ✅ ✅
XiaomiMiMo/MiMo-VL-7B-RL MiMo-VL (7B) ✅ ✅
AI-ModelScope/llava-v1.6-34b LLaVA (v1.5 & v1.6) ✅ ✅
lmms-lab/llava-next-72b LLaVA-NeXT (8B, 72B) ✅ ✅
moonshotai/Kimi-VL-A3B-Instruct Kimi-VL (A3B) ✅ ✅
ZhipuAI/GLM-4.5V GLM-4.5V (106B) ✅ ✅
LLM-Research/Llama-3.2-11B-Vision-Instruct Llama 3.2 Vision (11B) ✅ ✅
rednote-hilab/dots.ocr DotsVLM-OCR ✅ ✅
Qwen/Qwen3-Omni-30B-A3B-Instruct Qwen3-Omni ✅ ✅
stepfun-ai/Step3-VL-10B Step3-VL (10B) ✅ ✅

임베딩 모델 (Embedding Models)

모델 모델 패밀리 A2 지원 A3 지원
intfloat/e5-mistral-7b-instruct E5 (Llama/Mistral based) ✅ ✅
iic/gte-Qwen2-1.5B-instruct GTE-Qwen2 ✅ ✅
Qwen/Qwen3-Embedding-8B Qwen3-Embedding ✅ ✅
iic/gme-Qwen2-VL-2B-Instruct GME (Multimodal) ✅ ✅
AI-ModelScope/clip-vit-large-patch14-336 CLIP ✅ ✅
BAAI/bge-large-en-v1.5 BGE ✅ ✅

보상 모델 (Reward Models)

모델 모델 패밀리 A2 지원 A3 지원
Skywork/Skywork-Reward-Llama-3.1-8B-v0.2 Llama3.1 Reward ✅ ✅
Shanghai_AI_Laboratory/internlm2-7b-reward InternLM 2 Reward ✅ ✅
Qwen/Qwen2.5-Math-RM-72B Qwen2.5 Reward - Math ✅ ✅
Howeee/Qwen2.5-1.5B-apeach Qwen2.5 Reward - Sequence ✅ ✅

재랭크 모델 (Rerank Models)

모델 모델 패밀리 A2 지원 A3 지원
BAAI/bge-reranker-v2-m3 BGE-Reranker ✅ ✅
Qwen/Qwen3-Reranker-8B Qwen3-Reranker (decoder-only yes/no) ✅ ✅
Qwen/Qwen3-VL-Reranker-2B Qwen3-VL-Reranker (multimodal yes/no) ✅ ✅

더 알아보기 (Learn more)