- 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
Qwen3.8-Flash-Next
Alibaba (Qwen)
Apache 2.0
Open language model for chat and text. 3 sizes from 27.8B to 2.4T parameters, partly mixture-of-experts (only a part of the parameters is active per token).
- Mixture-of-experts: 180.0B parameters in total, of which 61.6B 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× Tesla V100 32 GB
The model needs about 109.7 GB of memory. Passively cooled, fan and 8-pin EPS cable required.
- Weights
- 108.0 GB
- KV cache
- 0.2 GB
- Speed
- ~10.7 Tok./s
More suitable configurations
| Card | Quantity | Tok./s | Status |
|---|---|---|---|
| Tesla T4 16 GB | 8× | 3.1 | 2-3 weeks delivery time |
| Tesla P100 16 GB | 8× | 3.9 | 2-3 weeks delivery time |
| Tesla P40 24 GB | 8× | 3.6 | 2-3 weeks delivery time |
| Tesla V100 16 GB | 8× | 8.5 | only 1 in stock |
| Tesla V100 32 GB | 4× | 10.7 | On order, delivery time approx. 1 week |
| AMD Instinct MI50 16 GB | 8× | 6.5 | only 2 in stock |
| AMD Instinct MI50 32 GB | 4× | 8.2 | only 2 in stock |
| AMD Instinct MI60 32 GB | 4× | 7.6 | 2-3 weeks delivery time |
| AMD Instinct MI100 32 GB | 4× | 9.2 | 2-3 weeks delivery time |
| A30 24 GB | 8× | 8.3 | 2-3 weeks delivery time |
| RTX A5000 24 GB | 8× | 8.6 | 2-3 weeks delivery time |
| A40 48 GB | 4× | 10.3 | 2-3 weeks delivery time |
