Can you run Mistral Large 3 (675B) locally?
Mistral family · Dec 2025 · 675B parameters (41B activated per token) · Hugging Face model card
Mistral Large 3 is a multimodal granular mixture-of-experts model with open weights: 675B total parameters (673B language model + 2.5B vision encoder), 41B active per token, and a 256k-token context window, Apache-2.0 licensed. Running it locally needs roughly 410 GB of memory at Q4-class quantization (calculated) - multi-GPU server territory. Measured local speeds: not yet published here; we list unknowns as unknowns.
Minimum: 410 GB+ memory (Q4-class, calculated) · Recommended: 512 GB+ for comfortable context headroom (Q4-class, calculated)
How much VRAM does Mistral Large 3 (675B) need at each quantization?
Mistral Large 3 (675B) needs 445.5 GB of VRAM at Q4_K_M. The table below lists weights-only size and total VRAM including overhead for each common quantization level with 4k and 32k token contexts.
| Quantization | Bits / weight | Weights only | Total + KV @4k ctx | Total + KV @32k ctx |
|---|---|---|---|---|
| Q4_K_M | 4.8 | 405.0 GB | 445.5 GB | 445.5 GB |
| Q5_K_M | 5.7 | 480.9 GB | 529.0 GB | 529.0 GB |
| Q6_K | 6.6 | 556.9 GB | 612.6 GB | 612.6 GB |
| Q8_0 | 8.5 | 717.2 GB | 788.9 GB | 788.9 GB |
| FP16 | 16 | 1,350.0 GB | 1,485.0 GB | 1,485.0 GB |
Method: weights = parameters × bits-per-weight (Q4_K_M ≈ 4.8, Q5_K_M ≈ 5.7, Q6_K ≈ 6.6, Q8_0 ≈ 8.5, FP16 = 16), plus 10% loading overhead, plus KV cache from the verified architecture config (not applicable to this architecture). Source: vendor-card-verified 2026-09-12 (HF model card: 675B total / 41B active granular MoE, 256k context, Apache-2.0); VRAM figures are documented calculations, not measurements
Which GPUs can run Mistral Large 3 (675B) locally?
At Q4_K_M with a 4k context, Mistral Large 3 (675B) needs 445.5 GB of VRAM. The lists below are computed live from our GPU database and grouped by how much headroom the card has. Disclosure: CompareAIHardware.com participates in the Amazon Associates program and earns from qualifying purchases through links on this page. Affiliate relationships do not influence our recommendations.
Needs 2+ GPUs to run Mistral Large 3 (675B)
One of these cards is too small on its own, but a pair (tensor or pipeline parallel, ~90% efficiency) covers the 445.5 GB requirement. See our multi-GPU guide for setup.
| GPU | VRAM | Bandwidth | Type | |
|---|---|---|---|---|
| AMD Instinct MI325X | 256 GB ×2 | 6000 GB/s | Data Center GPU | |
| AMD Instinct MI355X | 288 GB ×2 | 8000 GB/s | Data Center GPU | |
| NVIDIA B300 (Blackwell Ultra) | 288 GB ×2 | 8000 GB/s | Data Center GPU |