Mistral Small 4 119B
MoEToolsVisionPopularMistral AI · 119B (6.5B active) · Mixture of Experts
Sparse Mistral Small 4 — 6.5B active, strong local all-rounder
0
0
2026-03
256K tokens
Use cases
Your hardware
16 GB Reference Desktop
Quantization options
What is quantization?| Quant | Bits | VRAM | Disk | Quality | Status |
|---|---|---|---|---|---|
| Q2_K | 2 | 38.6 GB | 36.4 GB | low | Too heavy · ~1 t/s |
| Q3_K_M | 3 | 53.8 GB | 50.9 GB | moderate | Too heavy · ~1 t/s |
| Q4_K_M | 4 | 61.5 GB | 58.2 GB | good | Too heavy · ~1 t/s |
| Q5_K_M | 5 | 76.7 GB | 72.7 GB | good | Too heavy · ~1 t/s |
| Q6_K | 6 | 91.9 GB | 87.3 GB | excellent | Too heavy · ~1 t/s |
| Q8_0 | 8 | 122.4 GB | 116.4 GB | excellent | Too heavy · ~1 t/s |
| F16 | 16 | 244.3 GB | 232.7 GB | lossless | Too heavy · ~1 t/s |
About this model
Mistral Small 4 119B (Mistral small 4) is a 119B sparse mixture-of-experts (MoE) model activating 6.5B active per token from Mistral AI in the Mistral family, designed primarily for chat, vision, code, reasoning workflows. Supports structured tool calling; use Q4_K_M or higher when strict JSON schema adherence matters. Supports up to 256K context tokens; start at 4K–8K with Q8_0 KV cache on tight GPUs.
Hardware planning baselines start at 66.5 GB minimum memory and 110.8 GB recommended memory. At Q4_K_M, weights plus baseline runtime overhead use roughly 61.5 GB of RAM/VRAM and 58.2 GB of disk space. At 119B, Mistral Small 4 119B belongs firmly in the workstation and multi-GPU tier. Running it comfortably requires 110.8 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 Mistral Small 4 119B locally via llama.cpp with Flash Attention enabled. Under Apache License 2.0, review official weight cards before commercial redistribution.
Highlights
- Local Fit Baseline66.5 GB minimum and 110.8 GB recommended RAM/VRAM.
- Primary QuantQ4_K_M targets 61.5 GB memory and 58.2 GB disk.
- Context & Capabilities256K max context · Sparse MoE · Tool calling.
System requirements by device
Memory planning targets 66.5 GB minimum RAM/VRAM and 110.8 GB recommended memory. The baseline quant is Q4_K_M, estimated at 61.5 GB memory and 58.2 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 61.5 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 (66.5 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 | 38.6 GB | 36.4 GB | low |
| Q3_K_M | 53.8 GB | 50.9 GB | moderate |
| Q4_K_M | 61.5 GB | 58.2 GB | good |
| Q5_K_M | 76.7 GB | 72.7 GB | good |
| Q6_K | 91.9 GB | 87.3 GB | excellent |
| Q8_0 | 122.4 GB | 116.4 GB | excellent |
| F16 | 244.3 GB | 232.7 GB | lossless |
How to run Mistral Small 4 119B 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 Mistral Small 4 119B on Multi-GPU or High-RAM Desktop
- Plan for 66.5 GB minimum combined memory and 110.8 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 Mistral Small 4 119B 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 61.5 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%.
Mistral Small 4 119B vs alternatives
| Model | Params | Context | Min RAM | License |
|---|---|---|---|---|
| Mistral Small 4 119B | 119B (6.5B active) | 256K | 66.5 GB | Apache 2.0 |
| Devstral 2 123B | 123B | 256K | 68.7 GB | MRL |
| Mixtral 8x22B | 141B (39B active) | 64K | 78.8 GB | Apache 2.0 |
| 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 |
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 Mistral Small 4 119B?
- Mistral Small 4 119B is a 119B sparse mixture-of-experts model (6.5B active active per token) from Mistral AI, supporting up to 256K tokens with 66.5 GB minimum RAM and 110.8 GB recommended RAM.
- How do I download and install Mistral small 4?
- Download a GGUF build for Mistral Small 4 119B, start with Q4_K_M, and confirm the model fits locally before increasing context or quant quality.
- How much VRAM or RAM does Mistral Small 4 119B require?
- At Q4_K_M, Mistral Small 4 119B occupies roughly 61.5 GB of active memory and 58.2 GB on disk. Add 2–6 GB of spare headroom for KV cache buffers, runtime scratch space, and OS apps (110.8 GB recommended).
- What is the best way to run Mistral Small 4 119B locally?
- Run Mistral Small 4 119B in llama.cpp starting with Q4_K_M. Keep initial context at 4K–8K tokens and confirm 100% GPU/Metal offload before scaling context length.
- Can Mistral Small 4 119B run on a consumer laptop or smartphone?
- No smartphone or 8–16 GB laptop can hold Mistral Small 4 119B cleanly in local memory. It targets multi-GPU NVIDIA workstations, high-RAM Mac Studios, or MoE CPU-RAM hybrid rigs with at least 66.5 GB total memory.
- How does Mistral Small 4 119B compare to Devstral 2 123B?
- Devstral 2 123B (123B, 68.7 GB min RAM) sits closest in size. Compare active parameters, context ceiling (256K vs 256K), and license terms (Apache 2.0) before switching.