• 3 % discount on prepayment
  • Insured shipping from Switzerland
  • Sustainable: refurbished instead of new
  • Warranty on all cards
← Back to overview

GLM-4.7-Flash

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: 31.2B parameters in total, of which 5.2B 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
4k128k
How much text the model keeps in view at once. More context needs more memory.
Recommendation
1× H100 80 GB

The model needs about 20.7 GB of memory. Hopper, FP8 support.

Weights
18.7 GB
KV cache
0.4 GB
Speed
~306.1 Tok./s

More suitable configurations

CardQuantityTok./sStatus
Tesla T4 16 GB2×49.02-3 weeks delivery time
Tesla P100 16 GB2×61.82-3 weeks delivery time
Tesla P40 24 GB1×62.52-3 weeks delivery time
Tesla V100 16 GB2×136.6only 1 in stock
Tesla V100 32 GB1×147.5On order, delivery time approx. 1 week
AMD Instinct MI50 16 GB2×104.7in stock
AMD Instinct MI50 32 GB1×113.1in stock
AMD Instinct MI60 32 GB1×104.42-3 weeks delivery time
AMD Instinct MI100 32 GB1×126.32-3 weeks delivery time
A30 24 GB1×142.72-3 weeks delivery time
RTX A5000 24 GB1×148.72-3 weeks delivery time
A40 48 GB1×141.82-3 weeks delivery time