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

Llama-4-Scout-17B-16E-Instruct

Meta
GermanLlama 4 Community

Open language model for chat and text. 2 sizes from 108.6B to 401.6B parameters, partly mixture-of-experts (only a part of the parameters is active per token).

  • Mixture-of-experts: 108.6B parameters in total, of which 17.0B 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
4k256k
How much text the model keeps in view at once. More context needs more memory.
Recommendation
1× H100 80 GB

The model needs about 68.3 GB of memory. Hopper, FP8 support.

Weights
65.2 GB
KV cache
1.6 GB
Speed
~93.0 Tok./s

More suitable configurations

CardQuantityTok./sStatus
Tesla T4 16 GB8×10.32-3 weeks delivery time
Tesla P100 16 GB8×13.02-3 weeks delivery time
Tesla P40 24 GB4×15.32-3 weeks delivery time
Tesla V100 16 GB8×28.7only 1 in stock
Tesla V100 32 GB4×36.1On order, delivery time approx. 1 week
AMD Instinct MI50 16 GB8×22.0only 2 in stock
AMD Instinct MI50 32 GB4×27.7only 2 in stock
AMD Instinct MI60 32 GB4×25.62-3 weeks delivery time
AMD Instinct MI100 32 GB4×30.92-3 weeks delivery time
A30 24 GB4×35.02-3 weeks delivery time
RTX A5000 24 GB4×36.42-3 weeks delivery time
A40 48 GB2×39.92-3 weeks delivery time