- 3 % discount on prepayment
- Insured shipping from Switzerland
- Sustainable: refurbished instead of new
- Warranty on all cards
- 3 % discount on prepayment
- Insured shipping from Switzerland
- Sustainable: refurbished instead of new
- Warranty on all cards
← Back to overview
NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16
NVIDIA
GermanOpenMDW 1.1
Open language model for chat and text. 31.6B Parameter, partly mixture-of-experts (only a part of the parameters is active per token).
- Mixture-of-experts: 31.6B parameters in total, of which 3.0B active per token. Memory requirement depends on all parameters, speed on the active ones.
Settings
Quality
4 bit (Q4_K_M) is the most common choice: good quality with low memory requirements.
Context length: 8k tokens
4k256k
How much text the model keeps in view at once. More context needs more memory.
Recommendation
1× H100 80 GB
The model needs about 20.9 GB of memory. Hopper, FP8 support.
- Weights
- 19.0 GB
- KV cache
- 0.4 GB
- Speed
- ~506.9 Tok./s
More suitable configurations
| Card | Quantity | Tok./s | Status |
|---|---|---|---|
| Tesla T4 16 GB | 2× | 81.2 | 2-3 weeks delivery time |
| Tesla P100 16 GB | 2× | 102.3 | 2-3 weeks delivery time |
| Tesla P40 24 GB | 1× | 103.6 | 2-3 weeks delivery time |
| Tesla V100 16 GB | 2× | 226.2 | only 1 in stock |
| Tesla V100 32 GB | 1× | 244.3 | On order, delivery time approx. 1 week |
| AMD Instinct MI50 16 GB | 2× | 173.4 | in stock |
| AMD Instinct MI50 32 GB | 1× | 187.3 | in stock |
| AMD Instinct MI60 32 GB | 1× | 172.9 | 2-3 weeks delivery time |
| AMD Instinct MI100 32 GB | 1× | 209.1 | 2-3 weeks delivery time |
| A30 24 GB | 1× | 236.4 | 2-3 weeks delivery time |
| RTX A5000 24 GB | 1× | 246.3 | 2-3 weeks delivery time |
| A40 48 GB | 1× | 234.9 | 2-3 weeks delivery time |
