RTX 5090 (32 GB) vs RTX 4090 (24 GB)
The RTX 5090 (32 GB) runs 86 of 110 open models at 4-bit or better, the RTX 4090 (24 GB) runs 80. Usable memory: about 31 GB against 23 GB. The RTX 5090 (32 GB) has more memory bandwidth (1792 vs 1008 GB/s), so models that fit both answer faster on it.
Data as of 2 November 2026. Speeds are estimates unless marked measured.
Running gpt-oss-20b
Runs
~261 tok/s
MXFP4 · 15.1 of 31 GB · ModelFit est.
Runs
~169 tok/s
MXFP4 · 15.1 of 23 GB · ModelFit est.
Single-user decode speed, estimated. Fastest machines for this model
Memory
32 GB
memory
24 GB
memory
~31 GB
usable for a model
~23 GB
usable for a model
1792 GB/s
memory bandwidth
1008 GB/s
memory bandwidth
What it runs
86
of 110 models run at 4-bit or better
80
of 110 models run at 4-bit or better
Our pick
Q4_K_M · ~346 tok/s ModelFit est.
Q4_K_M · ~41 tok/s ModelFit est.
Largest current model that fits
35B at Q4_K_M
30.7B at Q4_K_M
Popular models
Too big
needs ~68 GB
Too big
needs ~68 GB
Runs
MXFP4 · ~261 tok/s
Runs
MXFP4 · ~169 tok/s
Runs
Q4_K_M · ~346 tok/s
Too big
needs ~24 GB
Runs
Q8_0 · ~36 tok/s
Runs
Q4_K_M · ~51 tok/s
Runs
Q8_0 · ~195 tok/s
Runs
Q4_K_M · ~168 tok/s
Too big
needs ~44 GB
Too big
needs ~44 GB
Too big
needs ~67 GB
Too big
needs ~67 GB
Too big
needs ~177 GB
Too big
needs ~177 GB
Rent one by the hour
Use Change under a machine to pick another type, producer, model or variant. How the numbers are worked out