can it run
No. 12GB of memory can't hold the 33GB working set this model needs. A smaller quant or a bigger machine is required.
weights 26GB + KV cache 0.5GB at 4k context vs ~6GB usable → no. 26.5 GB vs 6 GB usable — does not fit
computed roofline: 336 GB/s × 0.48–0.72 efficiency window / 3.9GB (Q4_K_M) — bands, never points.
MoE: file is 26GB (that decides fit) but only ~1.05B of experts activate per token (~3.9GB streamed) — band below is the active-expert estimate, optimistic; community MoE rows run several× the naive bandwidth math
| model file | 26GB (GGUF Q4_K_M) |
| minimum memory | 33GB working set |
| this machine | 12GB GDDR6, 336 GB/s |
| params | 8x7B MoE |
| license | apache-2.0 |
| price | $375 |
easy run: ollama pull mixtral:8x7b
runs with: ollama · llama.cpp · koboldcpp · lm studio
→ Mac mini M5 Pro (64GB) — cheapest machine that runs it ($1669)
→ every model the RTX 2060 12GB (used) can run — full list
→ rtx-5060-ti-16 — the next step up from this machine's GPU