back

Mixtral 8x22B

Apache 2.0

Mistral AI · 141B (39B active) · Mixture of Experts

Large MoE with 39B active params

16.5K downloads 746 likes 2024-04 64K context

Use Cases

chat code reasoning

Mixture of Experts

Total experts: 8
Active experts: 2
Active params: 39.1B

Quantization Options

Quant Bits VRAM Quality Status
Q2_K 2 45.6 GB low
Q3_K_M 3 63.7 GB moderate
Q4_K_M 4 72.7 GB good
Q5_K_M 5 90.8 GB good
Q6_K 6 108.8 GB excellent
Q8_0 8 144.9 GB excellent
F16 16 289.4 GB lossless

About this model

The Mixtral large Language Models (LLM) are a set of pretrained generative Sparse Mixture of Experts.

Sizes

  • mixtral:8x22b
  • mixtral:8x7b

Mixtral 8x22b

ollama run mixtral:8x22b

Mixtral 8x22B sets a new standard for performance and efficiency within the AI community. It is a sparse Mixture-of-Experts (SMoE) model that uses only 39B active parameters out of 141B, offering unparalleled cost efficiency for its size.

Mixtral 8x22B comes with the following strengths:

  • It is fluent in English, French, Italian, German, and Spanish
  • It has strong maths and coding capabilities
  • It is natively capable of function calling
  • 64K tokens context window allows precise information recall from large documents

References

Announcement

HuggingFace