Moore Threads MTT S4000
Datacenter
VRAM
48 GB
Memory
GDDR6
Bandwidth
768 GB/s
TDP
-
Suitable Workloads
Medium Language Models
Inference for models up to 70B parameters
Enterprise Deployment
Designed for 24/7 datacenter operations
Key Highlights
- Built on MUSA 3 architecture
- PCIe form factor
Memory
VRAM48 GB
Memory TypeGDDR6
Memory Bandwidth768 GB/s
Compute Performance
FP32 (Single Precision)25 TFLOPS
FP16 (Half Precision)100 TFLOPS
BF16100 TFLOPS
Architecture
ArchitectureMUSA 3
Release Year2023
Power & Physical
Form FactorPCIe
PCIe GenerationPCIe 5.0
Interconnect
Multi-GPU SupportYes
Notes
China market. Vendor-published specs: 100 TFLOPS FP16/BF16, 200 TOPS INT8, 48GB GDDR6 at 768 GB/s, MTLink multi-card interconnect. Used domestically for LLM training clusters.
Models That May Fit on Moore Threads MTT S4000
Estimates based on INT8 quantization. Actual fit depends on framework and batch size.
Explore More
Added Jul 30, 2026
Last updated: Jul 30, 2026
Your AI stack shouldn't stand still.
Every month new models become cheaper, faster, and more capable. Inferbase ensures your application automatically benefits without changing a single API call.