Mistral Large 3 675B-A41B — VRAM, speed & local setup
Written by Jakub Rusinowski · Last updated
PREVIEW — unverified. ~675B-total MoE with ~41B active per token, multimodal, Apache 2.0, 256K context. At Q4 it needs ~400 GB — an 8×80 GB server or equivalent. Apache 2.0 would make it one of the most permissive frontier-scale open weights if confirmed. Verify against the Hugging Face model card.
Mistral Large 3 675B-A41B needs about 408 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 Weights plus overhead only: this model's architecture is unpublished, so no KV cache is added.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 | 222.7 GB | — | Too big | |
| Q3_K_M 3.41 bpw | 288.5 GB | — | Too big | |
| Q4_K_M 4.83 bpw | 408.3 GB | — | Too big | |
| Q5_K_M 5.67 bpw | 479.2 GB | — | Too big | |
| Q6_K 6.56 bpw | 554.3 GB | — | Too big | |
| Q8_0 8.50 bpw | 718 GB | — | Too big | |
| F16 16.00 bpw | 1350.8 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 Large 3 675B-A41B VRAM calculator →
Get Mistral Large 3 675B-A41B running
Mistral Large 3 675B-A41B is not practical on a single consumer GPU: it needs about 408.3 GB at Q4_K_M, more than any single device in our catalogue holds with headroom. How multi-GPU inference works, or rent a multi-GPU server by the hour:
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 Large 3 675B-A41B
Install Ollama, then run:
ollama run mistral-large-3Specifications
Preview — The model is released, but these specs are thin or rest on a single source. Individual fields may be wrong.
Preview — The model is released, but these specs are thin or rest on a single source. Individual fields may be wrong.
- Parameters
- 675 Billion (41B active)
- Context window
- 256,000
- Architecture
- Mixture-of-Experts (multimodal)
- Provider
- Mistral AI
- Licence
- Apache 2.0
- Practical deployment
- Multi-GPU server
- Specified at
- Q4_K_M
- System RAM
- 1024 GB
- Record updated
- 2026-06-26
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 Large 3 675B-A41B on my GPU?
Mistral Large 3 675B-A41B — frequently asked questions
How much VRAM does Mistral Large 3 675B-A41B need?
About 408 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 Large 3 675B-A41B run on an RTX 4090 (24 GB)?
No. Mistral Large 3 675B-A41B needs about 408 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 Large 3 675B-A41B locally?
Install Ollama and run `ollama run mistral-large-3`. That pulls the weights and starts a local OpenAI-compatible endpoint; after the download nothing leaves the machine.