RTX 5090 gpt-oss-20B token-gen @128K context (Hardware Corner)
112.0 tok/s
Fields
id
benchmarks/comm-rtx-5090-gpt-oss-20b-tg-128kMetric
decode_tok_s
Value
112
Unit
tok/s
Model
gpt-oss-20B
Methodology
THIRD-PARTY MEASURED (Hardware Corner llama.cpp llama-bench). Q4_K_XL GGUF, CUDA, batch-1 decode at 128K context; 298.2 t/s @4K on the same card/model (hc-rtx5090-gpt-oss-20b-q4kxl-4k-tg128).
Measured by
third_party
Reproducible
yes
notes
32GB GDDR7, 1,792 GB/s. Same 128K number as the 120B-MXFP4 row (comm-rtx-5090-gpt-oss-120b-mxfp4-tg-128k) is a coincidence of the table, not a copy error - both rows are in the source.
Notes
32GB GDDR7, 1,792 GB/s. Same 128K number as the 120B-MXFP4 row (comm-rtx-5090-gpt-oss-120b-mxfp4-tg-128k) is a coincidence of the table, not a copy error - both rows are in the source.