Only with CPU offload
Olmo 3.1 32B Instruct at Q4_K_M needs 20.0 GB but only 14.4 GB is addressable, so about 31% of the layers would stream from system RAM at roughly 60 GB/s. Expect around 5.5 tokens per second — usable for batch work, painful for chat.
The largest fully open model you can audit end to end. Q4 fits 24 GB, tightly.
Hardware requirements What hardware do I need to run Olmo 3.1 32B Instruct? →At Q4_K_M with an 8K context: a 32 GB GPU (RTX 5090) or a 32 GB Mac (M4). Minimum, recommended, high-quality and long-context tiers, with all 114 devices checked at every quantisation.
Fits instead: Olmo 3 7B Instruct (7.2 GB)
The VRAM budget
Quantisation ladder
| Quant | Weights | Total @ 8K | Max context | Tok/s | Quality | Fit |
|---|---|---|---|---|---|---|
| Q8_0 | 31.9 GB | 33.7 GB | — | ~1.8 | −0.1% ppl | 19.3 GB over |
| Q6_K | 24.6 GB | 26.4 GB | — | ~2.8 | −0.4% ppl | 12.0 GB over |
| Q5_K_M | 21.3 GB | 23.1 GB | — | ~3.7 | −0.8% ppl | 8.7 GB over |
| Q4_K_M | 18.1 GB | 20.0 GB | — | ~5.5 | −1.9% ppl | 5.6 GB over |
| Q3_K_M | 14.7 GB | 16.5 GB | — | ~12 | −5.4% ppl | 2.1 GB over |
| Q2_K | 12.6 GB | 14.4 GB | 7K | ~35 | −15% ppl | 0.0 GB over |
Quality is the published perplexity delta against f16 weights. Max context assumes an f16 KV cache; q8_0 roughly doubles it. This model interleaves sliding-window layers (4096 tokens, 16 of 64 layers global), which is why its cache barely grows with context.
How to run it
$ llama-server \
-hf allenai/Olmo-3.1-32B-Instruct:Q4_K_M \
-c 8192 -ngl 44
The engine underneath most of the others. Every knob is exposed. More on llama.cpp.
Related models on a Tesla P100 16 GB
| Model | Total | Tok/s est. | |
|---|---|---|---|
| Olmo 3 7B Instruct 7.3B · same family |
Q4_K_M · 7.2 GB | 108 | Runs great |
| Qwen3 32B 32.8B · similar size |
Q4_K_M · 21.0 GB | ~4.8 | CPU offload |
| DeepSeek-R1-Distill-Qwen 32B 32.8B · similar size |
Q4_K_M · 21.0 GB | ~4.8 | CPU offload |
| Qwen2.5-Coder 32B 32.8B · similar size |
Q4_K_M · 21.0 GB | ~4.8 | CPU offload |
| Nemotron 3.5 Lightning 30B-A3B 31.6B MoE · similar size |
Q4_K_M · 18.4 GB | ~27 | CPU offload |
| Gemma 4 31B 31.3B · similar size |
Q4_K_M · 20.2 GB | ~5.3 | CPU offload |
| LLM-jp 4 33B Thinking 33.2B · similar size |
Q4_K_M · 21.3 GB | ~4.6 | CPU offload |
Olmo 3.1 32B Instruct on other hardware
| Device | Total | Tok/s est. | |
|---|---|---|---|
| GeForce RTX 4060 Laptop 7.0 GB usable · 256 GB/s |
20.0 GB | ~2.6 | CPU offload |
| GeForce RTX 3060 12 GB 10.6 GB usable · 360 GB/s |
20.0 GB | ~3.4 | CPU offload |
| M4 · 16 GB 10.7 GB usable · 120 GB/s |
20.0 GB | — | Won't fit |
| GeForce RTX 4090 22.4 GB usable · 1008 GB/s |
20.0 GB | 34 | Tight fit |
| Radeon RX 7900 XTX 22.4 GB usable · 960 GB/s |
20.0 GB | 32 | Tight fit |
| GeForce RTX 5090 30.4 GB usable · 1792 GB/s |
20.0 GB | 60 | Runs great |
| M4 · 32 GB 24.0 GB usable · 120 GB/s |
20.0 GB | 3.7 | Runs great |
| M3 Max · 64 GB 48.0 GB usable · 400 GB/s |
20.0 GB | 12 | Runs great |