3. 模型支持列表
该版本支持的模型列表如下：
注意
下面列表提供的模型均已通过正式测试，并由当前版本的编译器编译生成。
3.1. 模型功能支持列表
模型类型 |
模型名 |
量化 |
编译 |
示例包 |
备注 |
|---|---|---|---|---|---|
Classification |
MobileNetV2 |
✓ |
✓ |
✓ |
|
Classification |
ResNet50 |
✓ |
✓ |
✓ |
|
Object Detection |
YOLOv5s |
✓ |
✓ |
✓ |
|
Object Detection |
YOLOv8m |
✓ |
✓ |
✓ |
|
Autonomous Driving |
YOLOP |
✓ |
✓ |
✓ |
|
Object Detection |
YOLOv12m |
✓ |
✓ |
✓ |
|
Pose Estimation |
YOLOv8m-pose |
✓ |
✓ |
✓ |
|
Segmentation |
YOLOv8m-seg |
✓ |
✓ |
✓ |
|
Classification |
EfficientNet |
✓ |
✓ |
✓ |
|
Classification |
ViT-Base |
✓ |
✓ |
✓ |
|
Object Detection |
YOLOv3 |
✓ |
✓ |
✓ |
|
Embedding |
BGE |
✓ |
✓ |
✓ |
|
Embedding |
GTE-Qwen2-1.5B |
✓ |
✓ |
✓ |
|
ASR |
Whisper-medium |
✓ |
✓ |
✓ |
|
Classification |
YOLOv8m-cls |
✓ |
✓ |
✓ |
|
Object Detection |
YOLOv11m |
✓ |
✓ |
✓ |
|
Object Detection |
YOLOv10m |
✓ |
✓ |
✓ |
|
Object Detection |
YOLOv5m-face |
✓ |
✓ |
✓ |
|
Object Detection |
YOLOv7 |
✓ |
✓ |
✓ |
|
Object Detection |
YOLOv9m |
✓ |
✓ |
✓ |
|
Object Detection |
YOLOX |
✓ |
✓ |
✓ |
|
LLM |
GPT-OSS-20B-A3B |
✓ |
✓ |
✓ |
|
OCR |
LPRNet |
✓ |
✓ |
✓ |
|
ASR |
Whisper-large-v3-turbo-0.8B |
✓ |
✓ |
✓ |
|
ASR |
SenseVoiceSmall |
✓ |
✓ |
✓ |
|
Embedding |
Qwen3-embedding-0.6B/4B/8B |
✓ |
✓ |
✓ |
|
Reranker |
Qwen3-Reranker-0.6B/4B/8B |
✓ |
✓ |
✓ |
|
ASR |
Qwen3-ASR |
✓ |
✓ |
✓ |
|
VLM |
Qwen3.5-9B/4B/2B |
✓ |
✓ |
✓ |
|
VLM |
Qwen3.6-35B-A3B |
✓ |
✓ |
✓ |
|
OCR |
GLM-OCR |
✓ |
✓ |
✓ |
|
TTS |
CosyVoice3-0.5B |
✓ |
✓ |
✓ |
|
ASR |
GLM-ASR-Nano-2512 |
✓ |
✓ |
✓ |
|
Object Detection |
YOLOv26m |
✓ |
✓ |
✓ |
|
VLM |
Qwen3.5/3.6-27B |
✓ |
✓ |
✓ |
|
Omni |
Qwen3-Omni |
✓ |
✓ |
X |
|
VLM |
Gemma4-26B-A4B |
✓ |
✓ |
✓ |
|
ASR |
CAM++ |
✓ |
✓ |
✓ |
|
ASR |
punc_ct-transformer |
✓ |
✓ |
✓ |
|
Backbone |
DINOv3 |
✓ |
✓ |
✓ |
|
Media Gen |
Z-Image-Turbo |
✓ |
✓ |
✓ |
|
OCR |
MinerU-2.5-Pro-2604-1.2B |
✓ |
✓ |
✓ |
|
TTS |
Qwen3-TTS-0.6B/1.7B |
✓ |
✓ |
✓ |
|
VLM |
Gemma4-e2B/e4B |
✓ |
✓ |
✓ |
|
VLM |
Gemma4-31B |
✓ |
✓ |
✓ |
|
ASR |
emotion2vec |
✓ |
✓ |
✓ |
|
Embedding |
SigLIP 2 |
✓ |
✓ |
✓ |
|
Embedding |
Qwen3-VL-embedding-2B/8B |
✓ |
✓ |
✓ |
|
Segmentation |
SAM 2 |
✓ |
✓ |
✓ |
|
LLM |
Qwen3.5-122B-A10B |
✓ |
✓ |
✓ |
|
LLM |
Qwen-agentworld-35B-A3B |
✓ |
✓ |
✓ |
|
TTS |
voxCPM2 |
✓ |
✓ |
X |
|
LLM |
HY-MT2-7B |
✓ |
✓ |
X |
|
Segmentation |
SAM 3 |
✓ |
✓ |
✓ |
|
LALM |
fun-audio-chat-8B |
✓ |
✓ |
✓ |
|
VLM |
Ornith-1-35B |
✓ |
✓ |
✓ |
|
VLM |
Laguna-S-2.1-118B-A8B |
✓ |
✓ |
✓ |
|
VLM |
MiniCPM-v-4.6 |
✓ |
✓ |
✓ |
|
OCR |
PP-OCRv5 |
✓ |
✓ |
✓ |
|
ASR |
FunASR 1.2.6 |
✓ |
✓ |
X |
|
VLM |
MiniCPM-v-4.5 |
✓ |
✓ |
✓ |
|
VLM |
Qwen3.8-27B |
✓ |
✓ |
✓ |
|
OCR |
ppOCR-VL1.6 |
✓ |
✓ |
✓ |
3.2. 模型推理能力适配表
说明：
表中 “支持1芯”、“支持2芯”、 “支持4芯” 表示软件层面适配的算力规格，即后摩M50芯片数，而非硬件产品物理固有的芯片搭载数量。
✓：表示支持。X：表示不支持。—：表示当前版本未完成适配验证。
模型类型 |
模型名 |
支持1芯 |
支持2芯 |
支持4芯 |
支持多batch |
|---|---|---|---|---|---|
Classification |
MobileNetV2 |
✓ |
— |
— |
✓ |
Classification |
ResNet50 |
✓ |
— |
— |
✓ |
Object Detection |
YOLOv5s |
✓ |
— |
— |
✓ |
Object Detection |
YOLOv8m |
✓ |
— |
— |
✓ |
Autonomous Driving |
YOLOP |
✓ |
— |
— |
✓ |
Object Detection |
YOLOv12m |
✓ |
— |
— |
✓ |
Pose Estimation |
YOLOv8m-pose |
✓ |
— |
— |
✓ |
Segmentation |
YOLOv8m-seg |
✓ |
— |
— |
✓ |
Classification |
EfficientNet |
✓ |
— |
— |
✓ |
Classification |
ViT-Base |
✓ |
— |
— |
✓ |
Object Detection |
YOLOv3 |
✓ |
— |
— |
✓ |
Embedding |
BGE |
✓ |
— |
— |
— |
Embedding |
GTE-Qwen2-1.5B |
✓ |
— |
— |
— |
ASR |
Whisper-medium |
✓ |
— |
— |
— |
Classification |
YOLOv8m-cls |
✓ |
— |
— |
✓ |
Object Detection |
YOLOv11m |
✓ |
— |
— |
✓ |
Object Detection |
YOLOv10m |
✓ |
— |
— |
✓ |
Object Detection |
YOLOv5m-face |
✓ |
— |
— |
✓ |
Object Detection |
YOLOv7 |
✓ |
— |
— |
✓ |
Object Detection |
YOLOv9m |
✓ |
— |
— |
✓ |
Object Detection |
YOLOX |
✓ |
— |
— |
✓ |
LLM |
GPT-OSS-20B-A3B |
✓ |
✓ |
— |
— |
OCR |
LPRNet |
✓ |
— |
— |
— |
ASR |
Whisper-large-v3-turbo-0.8B |
✓ |
— |
— |
— |
ASR |
SenseVoiceSmall |
✓ |
— |
— |
— |
Embedding |
Qwen3-embedding-0.6B/4B/8B |
✓ |
— |
— |
— |
Reranker |
Qwen3-Reranker-0.6B/4B/8B |
✓ |
— |
— |
— |
ASR |
Qwen3-ASR |
✓ |
— |
— |
— |
VLM |
Qwen3.5-9B/4B/2B |
✓ |
— |
— |
✓ |
VLM |
Qwen3.6-35B-A3B |
✓ |
✓ |
— |
X |
OCR |
GLM-OCR |
✓ |
— |
— |
— |
TTS |
CosyVoice3-0.5B |
✓ |
— |
— |
— |
ASR |
GLM-ASR-Nano-2512 |
✓ |
— |
— |
— |
Object Detection |
YOLOv26m |
✓ |
— |
— |
✓ |
VLM |
Qwen3.5/3.6-27B |
✓ |
✓ |
✓ |
X |
Omni |
Qwen3-Omni |
✓ |
— |
— |
— |
VLM |
Gemma4-26B-A4B |
✓ |
— |
— |
— |
ASR |
CAM++ |
✓ |
— |
— |
— |
ASR |
punc_ct-transformer |
✓ |
— |
— |
— |
Backbone |
DINOv3 |
✓ |
— |
— |
— |
Media Gen |
Z-Image-Turbo |
✓ |
— |
— |
— |
OCR |
MinerU-2.5-Pro-2604-1.2B |
✓ |
— |
— |
— |
TTS |
Qwen3-TTS-0.6B/1.7B |
✓ |
— |
— |
— |
VLM |
Gemma4-e2B/e4B |
✓ |
— |
— |
— |
VLM |
Gemma4-31B |
X |
✓ |
— |
— |
ASR |
emotion2vec |
✓ |
— |
— |
— |
Embedding |
SigLIP 2 |
✓ |
— |
— |
— |
Embedding |
Qwen3-VL-embedding-2B/8B |
✓ |
— |
— |
— |
Segmentation |
SAM 2 |
✓ |
— |
— |
— |
LLM |
Qwen3.5-122B-A10B |
X |
✓ |
✓ |
— |
LLM |
Qwen-agentworld-35B-A3B |
✓ |
✓ |
— |
— |
TTS |
voxCPM2 |
✓ |
— |
— |
— |
LLM |
HY-MT2-7B |
✓ |
✓ |
— |
— |
Segmentation |
SAM 3 |
✓ |
— |
— |
— |
LALM |
fun-audio-chat-8B |
✓ |
— |
— |
— |
VLM |
Ornith-1-35B |
✓ |
— |
— |
— |
VLM |
Laguna-S-2.1-118B-A8B |
— |
✓ |
— |
— |
VLM |
MiniCPM-v-4.6 |
✓ |
— |
— |
— |
OCR |
PP-OCRv5 |
✓ |
— |
— |
— |
ASR |
FunASR 1.2.6 |
✓ |
— |
— |
— |
VLM |
MiniCPM-v-4.5 |
✓ |
— |
— |
— |
VLM |
Qwen3.8-27B |
✓ |
✓ |
— |
— |
OCR |
ppOCR-VL1.6 |
✓ |
— |
— |
— |