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

DeepSeek-V4.1-Flash

DeepSeek
MIT

Open language model for chat and text. 3 sizes from 304.2B to 1.7T parameters, partly mixture-of-experts (only a part of the parameters is active per token).

  • Mixture-of-experts: 763.2B parameters in total, of which 228.1B 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 460.1 GB of memory. Hopper, FP8 support.

Weights
457.9 GB
KV cache
0.7 GB
Speed
~4.8 Tok./s

More suitable configurations

CardQuantityTok./sStatus
A100 80 GB8×3.42-3 weeks delivery time
H100 80 GB8×4.8only 1 in stock