Mistral Small 4 119B-A6.5B — VRAM, speed & local setup
Written by Jakub Rusinowski · Last updated
Unifies Mistral's reasoning, multimodal, and agentic-coding lines into one MoE checkpoint. 128 experts, 4 active per token (~6.5B active). At Q4_K_M it fits a single 24GB consumer GPU (RTX 4090/A10); FP8 and higher precision need 48GB+ workstation cards. 256K context, Apache 2.0.
Mistral Small 4 119B-A6.5B needs about 73 GB of VRAM at Q4_K_M — quantized weights plus framework overhead, before any KV cache. On Apple Silicon that figure comes out of unified memory.
VRAM and speed by quantization
Quoted against NVIDIA RTX 4090 (24 GB). Weights plus overhead only: this model's architecture is unpublished, so no KV cache is added.
| Quant | Memory | VRAM | Speed (est.) | Fit |
|---|---|---|---|---|
| Q2_K 2.63 bpw | 39.9 GB | ~11 tok/s | Offload | |
| Q3_K_M 3.41 bpw | 51.5 GB | ~9 tok/s | Offload | |
| Q4_K_M 4.83 bpw | 72.6 GB | — | Too big | |
| Q5_K_M 5.67 bpw | 85.1 GB | — | Too big | |
| Q6_K 6.56 bpw | 98.4 GB | — | Too big | |
| Q8_0 8.50 bpw | 127.2 GB | — | Too big | |
| F16 16.00 bpw | 238.8 GB | — | Too big |
Black marker = usable memory on the NVIDIA RTX 4090 (24 GB). Estimates from the memory-bandwidth roofline on the methodology page. Mistral Small 4 119B-A6.5B VRAM calculator →
Get Mistral Small 4 119B-A6.5B running
The cheapest catalogued GPU that runs Mistral Small 4 119B-A6.5B is the NVIDIA DGX Spark (128 GB).
or compare on Vast.ai from $0.77/hr (typical low · varies)
As an Amazon Associate we earn from qualifying purchases. Cloud GPU links are referral links — we may earn a commission at no extra cost to you.
How to run Mistral Small 4 119B-A6.5B
Install Ollama, then run:
ollama run mistral-smallSpecifications
Verified — Checked against the primary source — the model card or the vendor spec page — and corroborated by a second independent source.
- Parameters
- 119 Billion (~6.5B active)
- Context window
- 256,000
- Architecture
- Mixture-of-Experts (128 experts, 4 active)
- Provider
- Mistral AI
- Licence
- Apache 2.0
- Practical deployment
- Workstation (large unified memory or pro GPU)
- Specified at
- Q4_K_M
- System RAM
- 32 GB
- Record updated
- 2026-03-16
Commercial use permitted. No usage restrictions beyond attribution.
Quality and use cases
Scores as published by the model’s authors or an independent evaluator — quality, not throughput, and not measured by us.
Can I run Mistral Small 4 119B-A6.5B on my GPU?
- Mistral Small 4 on AMD Ryzen AI Max+ 395
- Mistral Small 4 on Apple M1 Max
- Mistral Small 4 on Apple M1 Pro
- Mistral Small 4 on Apple M1 Ultra
- Mistral Small 4 on Apple M2 Max
- Mistral Small 4 on Apple M2 Pro
- Mistral Small 4 on Apple M3 Max
- Mistral Small 4 on Apple M3 Pro
- Mistral Small 4 on Apple M4
- Mistral Small 4 on Apple M4 Max
- Mistral Small 4 on Apple M5
- Mistral Small 4 on Apple M5 Max
- Mistral Small 4 on Apple M5 Pro
- Mistral Small 4 on NVIDIA A100 80GB (PCIe)
- Mistral Small 4 on NVIDIA DGX Spark
- Mistral Small 4 on NVIDIA GeForce RTX 5090
- Mistral Small 4 on NVIDIA H100 80GB (PCIe)
- Mistral Small 4 on NVIDIA L40S
- Mistral Small 4 on NVIDIA RTX 6000 Ada Generation
- Mistral Small 4 on NVIDIA RTX PRO 6000 Blackwell
Mistral Small 4 119B-A6.5B — frequently asked questions
How much VRAM does Mistral Small 4 119B-A6.5B need?
About 73 GB at Q4_K_M — quantized weights plus framework overhead, before any KV cache. The cache grows with context length and is added on top; the table above folds it in. Apple Silicon counts unified memory toward the same figure.
Does Mistral Small 4 119B-A6.5B run on an RTX 4090 (24 GB)?
No. Mistral Small 4 119B-A6.5B needs about 73 GB at Q4_K_M, more than a single RTX 4090's 24 GB. It needs a larger card, several GPUs, or Apple Silicon with enough unified memory — or it runs with part of the weights offloaded to system RAM, which is much slower.
How do I run Mistral Small 4 119B-A6.5B locally?
Install Ollama and run `ollama run mistral-small`. That pulls the weights and starts a local OpenAI-compatible endpoint; after the download nothing leaves the machine.