- 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
MiniMax-M2.7
MiniMax
Eigene Lizenz
Open language model for chat and text. 2 sizes from 228.7B to 427B parameters, partly mixture-of-experts (only a part of the parameters is active per token).
- Mixture-of-experts: 228.7B parameters in total, of which 11.0B 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
4k128k
How much text the model keeps in view at once. More context needs more memory.
Recommendation
2× H100 80 GB
The model needs about 140.8 GB of memory. Hopper, FP8 support.
- Weights
- 137.2 GB
- KV cache
- 2.1 GB
- Speed
- ~124.0 Tok./s
More suitable configurations
| Card | Quantity | Tok./s | Status |
|---|---|---|---|
| Tesla P40 24 GB | 8× | 17.5 | 2-3 weeks delivery time |
| Tesla V100 32 GB | 8× | 41.4 | only 6 in stock |
| AMD Instinct MI50 32 GB | 8× | 31.7 | only 2 in stock |
| AMD Instinct MI60 32 GB | 8× | 29.3 | 2-3 weeks delivery time |
| AMD Instinct MI100 32 GB | 8× | 35.4 | 2-3 weeks delivery time |
| A30 24 GB | 8× | 40.0 | 2-3 weeks delivery time |
| RTX A5000 24 GB | 8× | 41.7 | 2-3 weeks delivery time |
| A40 48 GB | 4× | 50.0 | 2-3 weeks delivery time |
| RTX A6000 48 GB | 4× | 55.9 | 2-3 weeks delivery time |
| Quadro RTX 8000 48 GB | 4× | 41.4 | 2-3 weeks delivery time |
| L40 48 GB | 4× | 61.3 | 2-3 weeks delivery time |
| A100 40 GB | 4× | 61.9 | 2-3 weeks delivery time |
