- 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
GLM-5.3
Z.ai (Zhipu)
MIT
Open language model for chat and text. 3 sizes from 31.2B to 753.3B parameters, partly mixture-of-experts (only a part of the parameters is active per token).
- Mixture-of-experts: 753.3B parameters in total, of which 51.2B 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
8× H100 80 GB
The model needs about 454.2 GB of memory. Hopper, FP8 support.
- Weights
- 452.0 GB
- KV cache
- 0.7 GB
- Speed
- ~21.1 Tok./s
More suitable configurations
| Card | Quantity | Tok./s | Status |
|---|---|---|---|
| A100 80 GB | 8× | 15.0 | 2-3 weeks delivery time |
| H100 80 GB | 8× | 21.1 | only 1 in stock |
