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

Mistral-Medium-3.5-128B

Mistral AI
GermanApache 2.0

Open language model for chat and text. 3 sizes from 119.4B to 675B parameters, partly mixture-of-experts (only a part of the parameters is active per token).

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 81.1 GB of memory. Passively cooled, fan and 8-pin EPS cable required.

Weights
76.6 GB
KV cache
3.0 GB
Speed
~5.1 Tok./s

More suitable configurations

CardQuantityTok./sStatus
Tesla T4 16 GB8×1.52-3 weeks delivery time
Tesla P100 16 GB8×1.82-3 weeks delivery time
Tesla P40 24 GB4×2.22-3 weeks delivery time
Tesla V100 16 GB8×4.0only 1 in stock
Tesla V100 32 GB4×5.1On order, delivery time approx. 1 week
AMD Instinct MI50 16 GB8×3.1only 2 in stock
AMD Instinct MI50 32 GB4×3.9only 2 in stock
AMD Instinct MI60 32 GB4×3.62-3 weeks delivery time
AMD Instinct MI100 32 GB4×4.42-3 weeks delivery time
A30 24 GB4×4.92-3 weeks delivery time
RTX A5000 24 GB4×5.12-3 weeks delivery time
A40 48 GB2×5.62-3 weeks delivery time