Baidu Kunlun II
Kunlun Core Architecture
Active
Launched August 2021
Core Specifications
VendorBaidu
ArchitectureKunlun Core
Form Factor—
VRAM32 GB
Memory Bandwidth—
TDP200 W
Compute Performance
| Precision | TFLOPs |
|---|---|
| FP16 | 256 |
Performance Benchmarks
search ranking
| Configuration | Precision | Performance | Source |
|---|---|---|---|
| Baidu Search ranking models | INT8 | 85,000 queries_per_second | View |
llm inference
| Configuration | Precision | Performance | Source |
|---|---|---|---|
| Baidu Cloud inference optimization | FP16 | 4,000 tokens_per_second | View |
llm train
| Configuration | Precision | Performance | Source |
|---|---|---|---|
| ERNIE models, Baidu ecosystem | FP16 | 6,200 tokens_per_second | View |