Mixtral 8x7B
Mistral AIMixture-of-experts model that activates only 12.9B parameters per token, giving near-70B quality at far lower compute cost. Excellent for multi-task workloads.
Parameters 46.7B (8x7B MoE)
Min VRAM 24 GB
Recommended VRAM 32 GB
Context Length 32K
License Apache 2.0
🚀 Get Started
Run Mixtral 8x7B locally with one command:
ollama run mixtral Requires Ollama installed.
📊 Benchmarks
BenchmarkScore
MMLU 70.6
GSM8K 74.4
HumanEval 40.2
💻 Hardware Recommendations
🟢 Minimum
24 GB VRAM GPU or 48+ GB RAM (CPU mode)
Expect slower generation in CPU mode
🔵 Recommended
32 GB VRAM GPU
Fast generation with room for context