~/model/huggingfacemistralai/Mixtral-8x22B-Instruct-v0.1
Mistral provider

Mixtral 8x22B

Big Mixtral. 141B total, 39B active. Multi-GPU or M-series Ultra territory.

Parameters
141B
Family
Mistral
Context
64K tokens
FP16 weights
282GB
// where you can run it
mixtral:8x22bLM StudiovLLMMLXoMLX
// hugging face stats (cached daily)
26K downloads · 751 likes · license: apache-2.0 · updated 10 months ago

What you need to run this.

$ ./vrambudget --model mixtral-8x22b --by quant
FP16282 GB
M3 Ultra 512512 GB · fitsM2 Ultra 192192 GB · overB200192 GB · overMI300X192 GB · over
Q8_0150 GB
2× H100 NVL188 GB · fitsM2 Ultra 192192 GB · fitsB200192 GB · fitsMI300X192 GB · fits
Q6_K116 GB
H200141 GB · fits2× H100 NVL188 GB · fitsM2 Ultra 192192 GB · fitsB200192 GB · fits
Q5_K_M97 GB
M4 Max 128128 GB · fitsM5 Max 128128 GB · fitsDGX Spark128 GB · fitsGaudi 3128 GB · fits
Q4_K_M79 GB
M4 Max 128128 GB · fitsM5 Max 128128 GB · fitsDGX Spark128 GB · fitsGaudi 3128 GB · fits

// budgets shown at ctx 8K, concurrency 1, 15% safety headroom. Tune in the calculator →

Alternatives at this size.

$ grep --params similar catalog.json
$ similargpt-oss 120B117BOpenAI$ similarCommand R+104BCohere$ similarQwen 2.5 72B72.7BAlibaba$ similarLlama 3.3 70B70.6BMeta

Discussion.

$ gh discussion list

// sign in with github to leave a comment. threads live in the repo's discussions tab.