Only with CPU offload
Granite 4.1 8B at Q4_K_M needs 6.8 GB but only 5.3 GB is addressable, so about 30% of the layers would stream from system RAM at roughly 60 GB/s. Expect around 17 tokens per second — usable for batch work, painful for chat.
Matches the old Granite 4.0 32B MoE at a quarter of the size. Fast, Apache 2.0, no reasoning traces.
Hardware requirements What hardware do I need to run Granite 4.1 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: Granite 4.1 3B (3.1 GB)
The VRAM budget
Quantisation ladder
| Quant | Weights | Total @ 8K | Max context | Tok/s | Quality | Fit |
|---|---|---|---|---|---|---|
| F16 | 16.4 GB | 18.2 GB | — | ~2.7 | Reference | 12.9 GB over |
| Q8_0 | 8.7 GB | 10.5 GB | — | ~6.2 | −0.1% ppl | 5.2 GB over |
| Q6_K | 6.7 GB | 8.6 GB | — | ~9.4 | −0.4% ppl | 3.3 GB over |
| Q5_K_M | 5.8 GB | 7.7 GB | — | ~12 | −0.8% ppl | 2.4 GB over |
| Q4_K_M | 4.9 GB | 6.8 GB | — | ~17 | −1.9% ppl | 1.5 GB over |
| Q3_K_M | 4.0 GB | 5.9 GB | 4K | ~31 | −5.4% ppl | 0.6 GB over |
| Q2_K | 3.4 GB | 5.3 GB | 8K | 59 | −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 ibm-granite/granite-4.1-8b:Q4_K_M \
-c 8192 -ngl 27
The engine underneath most of the others. Every knob is exposed. More on llama.cpp.
Related models on a GeForce RTX 3060 Laptop
| Model | Total | Tok/s est. | |
|---|---|---|---|
| Granite 4.1 3B 3.4B · same family |
Q4_K_M · 3.1 GB | 106 | Runs great |
| Granite 4.1 30B 28.9B · same family |
Q4_K_M · 18.9 GB | ~2.6 | CPU offload |
| Ministral 3 8B 8.92B · similar size |
Q4_K_M · 6.7 GB | ~18 | CPU offload |
| LFM2.5 8B-A1B 8.47B MoE · similar size |
Q4_K_M · 5.5 GB | ~81 | CPU offload |
| Fara 7B 8.29B · similar size |
Q4_K_M · 5.7 GB | ~31 | CPU offload |
| Ornith 1.5 9B 9.41B · similar size |
Q4_K_M · 6.1 GB | ~22 | CPU offload |
| Qwen3 8B 8.19B · similar size |
Q4_K_M · 6.3 GB | ~22 | CPU offload |
Granite 4.1 8B on other hardware
| Device | Total | Tok/s est. | |
|---|---|---|---|
| GeForce RTX 4060 Laptop 7.0 GB usable · 256 GB/s |
6.8 GB | 31 | Tight fit |
| Arc B570 8.8 GB usable · 380 GB/s |
6.8 GB | 47 | Runs great |
| GeForce RTX 3060 12 GB 10.6 GB usable · 360 GB/s |
6.8 GB | 44 | Runs great |
| M4 · 16 GB 10.7 GB usable · 120 GB/s |
6.8 GB | 14 | Runs great |
| GeForce RTX 4090 22.4 GB usable · 1008 GB/s |
6.8 GB | 123 | Runs great |
| Radeon RX 7900 XTX 22.4 GB usable · 960 GB/s |
6.8 GB | 118 | Runs great |
| M3 Max · 64 GB 48.0 GB usable · 400 GB/s |
6.8 GB | 45 | Runs great |