Can an RTX 5090 run gpt-oss-20b?
Yes: an RTX 5090 runs gpt-oss-20b at 324.2 tok/s at Q4_K_M, measured on llama.cpp with a 166-token prompt in October 2026.
2 signed runs on llm-speed. Single-request decode speed in tokens per second; the short-chat workload unless noted. 1 of 2 runs were measured by llm-speed (@llmspeed); the other one is community-submitted (signed on the submitter's own machine, not measured by llm-speed, and not used for llm-speed verdicts). Runs measured by llm-speed are credited to @llmspeed. Verdicts and headline numbers use only llm-speed's own runs and fit tests.
Measured tok/s by quant and runtime
| Quant | Runtime | Build | Decode tok/s (best run) | Workload | Runs | Source |
|---|---|---|---|---|---|---|
| Q4_K_M | llama.cpp | commit 1268eda | 324.2 | chat-short | 1 | Measured by llm-speed@llmspeed |
| not reported | llama.cpp | not reported | 318.4 | chat-short | 1 | Community-submitted |
Fastest quant that fits: Q4_K_M on llama.cpp, 324.2 tok/s (measured by llm-speed; it fits because it ran on this hardware). Quants without a run are not listed.
VRAM: not recorded for these runs. The VRAM calculator gives an estimate (predicted, not measured).
Every measured row
| Run | Date (UTC) | Workload | Quant | Runtime / build | Prompt tokens | Output tokens | Decode tok/s | Prefill tok/s | TTFT ms | VRAM GiB | Source |
|---|---|---|---|---|---|---|---|---|---|---|---|
| r_dwp5fgdapmw | 2026-10-17 | chat-short | Q4_K_M | llama.cpp · commit 1268eda | 166 | 256 | 324.2 | 586.8 | 283 | – | Measured by llm-speed@llmspeed |
| r_b9ul-vxh9sc | 2026-07-01 | chat-short | – | llama.cpp · not reported | 0 | 253 | 318.4 | – | 173 | – | Community-submitted |
| r_dwp5fgdapmw | 2026-10-17 | chat-long | Q4_K_M | llama.cpp · commit 1268eda | 3188 | 1024 | 311.0 | 12638.4 | 252 | – | Measured by llm-speed@llmspeed |
| r_b9ul-vxh9sc | 2026-07-01 | chat-long | – | llama.cpp · not reported | 0 | 1015 | 301.6 | – | 239 | – | Community-submitted |
| r_dwp5fgdapmw | 2026-10-17 | agent-trace | Q4_K_M | llama.cpp · commit 1268eda | 3287 | 840 | 319.6 | 23887.0 | 89 | – | Measured by llm-speed@llmspeed |
| r_b9ul-vxh9sc | 2026-07-01 | agent-trace | – | llama.cpp · not reported | 3362 | 840 | 304.5 | 26907.5 | 78 | – | Community-submitted |
| r_dwp5fgdapmw | 2026-10-17 | concurrent-decode | Q4_K_M | llama.cpp · commit 1268eda | 1024 | 256 | 322.5 | – | – | – | Measured by llm-speed@llmspeed |
| r_b9ul-vxh9sc | 2026-07-01 | concurrent-decode | – | llama.cpp · not reported | 1024 | 256 | 302.8 | – | – | – | Community-submitted |
Questions
Can an RTX 5090 run gpt-oss-20b?
Yes: an RTX 5090 runs gpt-oss-20b at 324.2 tok/s at Q4_K_M, measured on llama.cpp with a 166-token prompt in October 2026.
How many tokens per second does gpt-oss-20b get on an RTX 5090?
Decode speed measured by llm-speed, by quantization and runtime: Q4_K_M on llama.cpp: 324.2 tok/s. Based on 1 run measured by llm-speed; single-request chat-short unless noted. 1 community-submitted run is listed on the page, labelled, and not counted here.
What is the fastest quant of gpt-oss-20b that fits on an RTX 5090?
Q4_K_M on llama.cpp was the fastest measured setting at 324.2 tok/s; it fit because it ran. Other quants are listed only where a run measured them.
How much VRAM does gpt-oss-20b use on an RTX 5090?
Memory use was not recorded for these runs. The VRAM calculator gives an estimate, not a measurement.