OpenAI on NVIDIA Blackwell
Can I run gpt-oss 120B-A5.1B on an RTX 5090?
Not at Q4_K_M — it needs 67.4 GB against 29.3 GB available. You would need 3 of these cards.
Won't fit
67.4 GB of 29.3 GB · 230%070 GB
Weights
66.0 GB
KV cache
0.6 GB
Runtime overhead
0.8 GB
Over the limit
38.1 GB
Short by 38.1 GB. You can run it with 15 of 36 layers on the RTX 5090 and the rest in system RAM, at roughly 21.5 tok/s — usable for batch work, painful for chat. A smaller quantisation or a shorter context is usually the better trade.
Generation21.5tok/s
Prompt processing8606tok/s
Max context0tokens
KV per 1K tokens0GB
Every quantisation of gpt-oss 120B-A5.1B on a RTX 5090
Highlighted row is the highest quality that still fits at 8K context.
| Quantisation | bpw | Weights | Total | Verdict | Max ctx | tok/s |
|---|---|---|---|---|---|---|
| FP16 / BF16 | 16.00 | 217.6 GB | 218.9 GB | Won't fit | — | 4.79 |
| INT8 / W8A8 | 8.50 | 115.3 GB | 116.7 GB | Won't fit | — | 9.93 |
| FP8 (E4M3) | 8.00 | 108.8 GB | 110.1 GB | Won't fit | — | 10.8 |
| Q6_K | 6.56 | 89.2 GB | 90.6 GB | Won't fit | — | 14.0 |
| Q5_K_M | 5.67 | 77.1 GB | 78.5 GB | Won't fit | — | 17.2 |
| Q5_K_S | 5.52 | 75.1 GB | 76.4 GB | Won't fit | — | 17.6 |
| Q4_K_M | 4.85 | 66.0 GB | 67.4 GB | Won't fit | — | 21.5 |
| Q4_K_S | 4.58 | 62.4 GB | 63.8 GB | Won't fit | — | 23.6 |
| Q4_0 | 4.55 | 62.0 GB | 63.4 GB | Won't fit | — | 23.7 |
| AWQ 4-bit | 4.25 | 59.4 GB | 60.7 GB | Won't fit | — | 24.6 |
| GPTQ 4-bit | 4.25 | 59.4 GB | 60.7 GB | Won't fit | — | 24.6 |
| MXFP4 | 4.25 | 59.4 GB | 60.7 GB | Won't fit | — | 24.6 |
| IQ4_XS | 4.25 | 58.0 GB | 59.3 GB | Won't fit | — | 26.3 |
| Q3_K_M | 3.91 | 53.4 GB | 54.7 GB | Won't fit | — | 29.6 |
| IQ3_M | 3.70 | 50.6 GB | 51.9 GB | Won't fit | — | 32.6 |
| IQ3_XXS | 3.06 | 41.9 GB | 43.3 GB | Won't fit | — | 47.8 |
| Q2_K | 2.63 | 36.1 GB | 37.5 GB | Won't fit | — | 71.8 |
| IQ2_XXS | 2.06 | 28.5 GB | 29.8 GB | Won't fit | — | 251 |
| IQ1_M | 1.75 | 24.3 GB | 25.7 GB | Runs comfortably | 59K | 351 |