- 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
DeepSeek-V4-Flash-0731
DeepSeek
MIT
Open language model for chat and text. 3 sizes from 304.2B to 1.7T parameters, partly mixture-of-experts (only a part of the parameters is active per token).
- Mixture-of-experts: 304.2B parameters in total, of which 33.6B 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
4× H100 80 GB
The model needs about 184.7 GB of memory. Hopper, FP8 support.
- Weights
- 182.5 GB
- KV cache
- 0.7 GB
- Speed
- ~40.2 Tok./s
More suitable configurations
| Card | Quantity | Tok./s | Status |
|---|---|---|---|
| Tesla V100 32 GB | 8× | 15.4 | only 6 in stock |
| AMD Instinct MI50 32 GB | 8× | 11.8 | only 2 in stock |
| AMD Instinct MI60 32 GB | 8× | 10.9 | 2-3 weeks delivery time |
| AMD Instinct MI100 32 GB | 8× | 13.2 | 2-3 weeks delivery time |
| A40 48 GB | 8× | 14.8 | 2-3 weeks delivery time |
| RTX A6000 48 GB | 8× | 16.6 | 2-3 weeks delivery time |
| Quadro RTX 8000 48 GB | 8× | 12.2 | 2-3 weeks delivery time |
| L40 48 GB | 8× | 18.1 | 2-3 weeks delivery time |
| A100 40 GB | 8× | 18.3 | 2-3 weeks delivery time |
| A100 80 GB | 4× | 28.6 | 2-3 weeks delivery time |
| H100 80 GB | 4× | 40.2 | only 1 in stock |
