NVIDIA B300 SXM
VRAM
288 GB
Memory
HBM3e
Bandwidth
8000 GB/s
TDP
1400W
Suitable Workloads
Large Language Models
Training and inference for models like GPT-4, Llama 70B+
Deep Learning Training
High-performance training for neural networks
Distributed Training
Multi-node training with fast interconnects
High-Throughput Inference
Optimized for batched inference workloads
Key Highlights
- Built on Blackwell Ultra architecture
- SXM form factor
Memory
Compute Performance
Architecture
Power & Physical
Interconnect
Notes
Blackwell Ultra. FP8 rate identical to B200 (5/10 PF dense/sparse, so FP16 sparse figure matches B200 convention); NVFP4 15 PF dense. Shipping since Jan 2026. Liquid-cooled in all form factors.
Models That May Fit on NVIDIA B300 SXM
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.