Mixtral 8x7B
MoEToolsMistral AI · 47B (12.9B active) · Mixture of Experts
MoE with 12.9B active params
772.5K
4.6K
2023-12
32K tokens
Use cases
Your hardware
16 GB Reference Desktop
Quantization options
What is quantization?| Quant | Bits | VRAM | Disk | Quality | Status |
|---|---|---|---|---|---|
| Q2_K | 2 | 15.5 GB | 14.4 GB | low | Barely runs · ~13 t/s |
| Q3_K_M | 3 | 21.6 GB | 20.1 GB | moderate | Barely runs · ~4 t/s |
| Q4_K_M | 4 | 24.6 GB | 23 GB | good | Barely runs · ~3 t/s |
| Q5_K_M | 5 | 30.6 GB | 28.7 GB | good | Barely runs · ~2 t/s |
| Q6_K | 6 | 36.6 GB | 34.5 GB | excellent | Barely runs · ~1 t/s |
| Q8_0 | 8 | 48.6 GB | 46 GB | excellent | Too heavy · ~1 t/s |
| F16 | 16 | 96.8 GB | 91.9 GB | lossless | Too heavy · ~1 t/s |
About this model
Mixtral 8x7B is a 47B sparse mixture-of-experts (MoE) model activating 12.9B active per token from Mistral AI in the Mistral family, designed primarily for chat, code workflows. Supports structured tool calling; use Q4_K_M or higher when strict JSON schema adherence matters. Supports up to 32K context tokens; start at 4K–8K with Q8_0 KV cache on tight GPUs.
Hardware planning baselines start at 26.3 GB minimum memory and 43.8 GB recommended memory. At Q4_K_M, weights plus baseline runtime overhead use roughly 24.6 GB of RAM/VRAM and 23 GB of disk space. At 47B, Mixtral 8x7B belongs firmly in the workstation and multi-GPU tier. Running it comfortably requires 43.8 GB of total memory across dual RTX GPUs, high-memory Apple Silicon (Mac Studio Ultra/Max), or MoE hybrid CPU-RAM offloading—mobile devices and standard 16 GB laptops are too constrained for practical speeds.
Deploy Mixtral 8x7B locally via Ollama (`mixtral:8x7b`), Ollama with Flash Attention enabled. Under Apache License 2.0, review official weight cards before commercial redistribution.
Highlights
- Local Fit Baseline26.3 GB minimum and 43.8 GB recommended RAM/VRAM.
- Primary QuantQ4_K_M targets 24.6 GB memory and 23 GB disk.
- Context & Capabilities32K max context · Sparse MoE · Tool calling.
System requirements by device
Memory planning targets 26.3 GB minimum RAM/VRAM and 43.8 GB recommended memory. The baseline quant is Q4_K_M, estimated at 24.6 GB memory and 23 GB disk. Actual runtime usage shifts with context length, KV cache precision, and OS background overhead.
| Device | Notes |
|---|---|
| Mac (Apple silicon) | Requires a 48 GB–192 GB Mac Studio or MacBook Pro Max/Ultra with enough unified wired budget for 24.6 GB. |
| Windows / Linux GPU | Use dual/multi-GPU CUDA splitting (--tensor-split) or MoE CPU expert offloading (-ot) paired with 32–128 GB DDR5 system RAM. |
| Phones & Compact Edge Boards | Not practical on phones (26.3 GB min RAM exceeds Android LMK and iOS Jetsam limits). Pick a 2B–4B model on mobile. |
Start with a short session and leave memory for the operating system. A file fitting on disk does not guarantee that it will load in memory.
Estimated memory by quantization
These are rough estimates, not tested minimum requirements. Check the actual file and runtime documentation; context, offloading and operating-system overhead can change the memory needed.
| Quant Tier | Estimated memory | Disk Size | Quality |
|---|---|---|---|
| Q2_K | 15.5 GB | 14.4 GB | low |
| Q3_K_M | 21.6 GB | 20.1 GB | moderate |
| Q4_K_M | 24.6 GB | 23.0 GB | good |
| Q5_K_M | 30.6 GB | 28.7 GB | good |
| Q6_K | 36.6 GB | 34.5 GB | excellent |
| Q8_0 | 48.6 GB | 46.0 GB | excellent |
| F16 | 96.8 GB | 91.9 GB | lossless |
How to run Mixtral 8x7B on your device
Start with a short prompt and the smallest quant that still gives the quality you need. A clean first run is more useful than loading a larger file with no memory headroom.
Run Mixtral 8x7B on Multi-GPU or High-RAM Desktop
- Plan for 26.3 GB minimum combined memory and 43.8 GB recommended memory for Q4_K_M.
- Use llama.cpp `-ot ".ffn_.*_exps.=CPU"` or `--tensor-split` across multiple GPUs so attention and KV cache remain in fast VRAM.
- Enable Flash Attention (`-fa on`) and Q8_0 KV cache (`-ctk q8_0 -ctv q8_0`) to prevent context buffers from exceeding GPU headroom.
- Avoid attempting to run workstation-scale weights on smartphones or 8 GB laptops; use a 2B–8B checkpoint on portable devices.
Run Mixtral 8x7B on High-Memory Apple Silicon Mac
- Verify you are on a 48 GB–192 GB Mac Studio or MacBook Pro Max/Ultra capable of holding 24.6 GB in wired unified memory.
- Launch via `llama-server` or LM Studio with MLX/Metal enabled and set initial context conservatively to 4K–8K.
- Monitor Activity Monitor Memory Pressure to confirm zero SSD swap thrashing during token generation.
- Step down to a lighter quantization (e.g., Q3_K_M or IQ4_XS) if total memory utilization exceeds 75%.
Mixtral 8x7B vs alternatives
| Model | Params | Context | Min RAM | License |
|---|---|---|---|---|
| Mixtral 8x7B | 47B (12.9B active) | 32K | 26.3 GB | Apache 2.0 |
| Devstral Small 2 24B | 24B | 256K | 13.4 GB | Apache 2.0 |
| Mistral Small 3.1 24B | 24B | 128K | 13.4 GB | Apache 2.0 |
| Ministral 3 14B | 14B | 256K | 7.8 GB | Apache 2.0 |
| Mistral Nemo 12B | 12B | 128K | 6.7 GB | Apache 2.0 |
| Ministral 8B | 8B | 32K | 4.5 GB | MRL |
Use this table to choose the smallest nearby model that still matches the job. A related model with lower RAM can be the better local choice when the current page fits only at the edge of available memory.
Frequently asked questions
- What is Mixtral 8x7B?
- Mixtral 8x7B is a 47B sparse mixture-of-experts model (12.9B active active per token) from Mistral AI, supporting up to 32K tokens with 26.3 GB minimum RAM and 43.8 GB recommended RAM.
- How much VRAM or RAM does Mixtral 8x7B require?
- At Q4_K_M, Mixtral 8x7B occupies roughly 24.6 GB of active memory and 23 GB on disk. Add 2–6 GB of spare headroom for KV cache buffers, runtime scratch space, and OS apps (43.8 GB recommended).
- What is the best way to run Mixtral 8x7B locally?
- Use `ollama run mixtral:8x7b` for a quick CLI/API setup, or load Q4_K_M in LM Studio or llama.cpp with Flash Attention enabled when you want explicit KV cache and layer control.
- Can Mixtral 8x7B run on a consumer laptop or smartphone?
- No smartphone or 8–16 GB laptop can hold Mixtral 8x7B cleanly in local memory. It targets multi-GPU NVIDIA workstations, high-RAM Mac Studios, or MoE CPU-RAM hybrid rigs with at least 26.3 GB total memory.
- How does Mixtral 8x7B compare to Devstral Small 2 24B?
- Devstral Small 2 24B (24B, 13.4 GB min RAM) sits closest in size. Compare active parameters, context ceiling (32K vs 256K), and license terms (Apache 2.0) before switching.