Only with CPU offload
Ministral 3 8B at Q4_K_M needs 6.7 GB but only 5.3 GB is addressable, so about 27% of the layers would stream from system RAM at roughly 60 GB/s. Expect around 18 tokens per second — usable for batch work, painful for chat.
Mistral's 8B with images in. Plain GQA, so budget more KV cache than Qwen3.5 9B at the same context.
Hardware requirements What hardware do I need to run Ministral 3 8B? →At Q4_K_M with an 8K context: a 10 GB GPU (Arc B570) or a 16 GB Mac (M4). Minimum, recommended, high-quality and long-context tiers, with all 114 devices checked at every quantisation.
Fits instead: Ministral 3 3B (3.6 GB)
The VRAM budget
Quantisation ladder
| Quant | Weights | Total @ 8K | Max context | Tok/s | Quality | Fit |
|---|---|---|---|---|---|---|
| F16 | 16.6 GB | 18.3 GB | — | ~2.7 | Reference | 13.0 GB over |
| Q8_0 | 8.8 GB | 10.5 GB | — | ~6.2 | −0.1% ppl | 5.2 GB over |
| Q6_K | 6.8 GB | 8.5 GB | — | ~9.5 | −0.4% ppl | 3.2 GB over |
| Q5_K_M | 5.9 GB | 7.6 GB | — | ~13 | −0.8% ppl | 2.3 GB over |
| Q4_K_M | 5.0 GB | 6.7 GB | — | ~18 | −1.9% ppl | 1.4 GB over |
| Q3_K_M | 4.1 GB | 5.7 GB | 4K | ~34 | −5.4% ppl | 0.4 GB over |
| Q2_K | 3.5 GB | 5.1 GB | 9K | 58 | −15% ppl | Fits |
Quality is the published perplexity delta against f16 weights. Max context assumes an f16 KV cache; q8_0 roughly doubles it.
How to run it
$ llama-server \
-hf mistralai/Ministral-3-8B-Instruct-2512:Q4_K_M \
-c 8192 -ngl 24
The engine underneath most of the others. Every knob is exposed. More on llama.cpp.
Related models on a GeForce GTX 1660 Super
| Model | Total | Tok/s est. | |
|---|---|---|---|
| Ministral 3 14B 13.9B · same family |
Q4_K_M · 9.7 GB | ~7.3 | CPU offload |
| Ministral 3 3B 3.85B · same family |
Q4_K_M · 3.6 GB | 94 | Runs great |
| Granite 4.1 8B 8.79B · similar size |
Q4_K_M · 6.8 GB | ~17 | CPU offload |
| LFM2.5 8B-A1B 8.47B MoE · similar size |
Q4_K_M · 5.5 GB | ~81 | CPU offload |
| Ornith 1.5 9B 9.41B · similar size |
Q4_K_M · 6.1 GB | ~22 | CPU offload |
| Fara 7B 8.29B · similar size |
Q4_K_M · 5.7 GB | ~31 | CPU offload |
| Qwen3.5 9B 9.65B · similar size |
Q4_K_M · 6.3 GB | ~21 | CPU offload |
Ministral 3 8B on other hardware
| Device | Total | Tok/s est. | |
|---|---|---|---|
| GeForce RTX 4060 Laptop 7.0 GB usable · 256 GB/s |
6.7 GB | 31 | Tight fit |
| Arc B570 8.8 GB usable · 380 GB/s |
6.7 GB | 46 | Runs great |
| GeForce RTX 3060 12 GB 10.6 GB usable · 360 GB/s |
6.7 GB | 43 | Runs great |
| M4 · 16 GB 10.7 GB usable · 120 GB/s |
6.7 GB | 13 | Runs great |
| GeForce RTX 4090 22.4 GB usable · 1008 GB/s |
6.7 GB | 122 | Runs great |
| Radeon RX 7900 XTX 22.4 GB usable · 960 GB/s |
6.7 GB | 116 | Runs great |
| M3 Max · 64 GB 48.0 GB usable · 400 GB/s |
6.7 GB | 45 | Runs great |