Can an RTX 4090 run gpt-oss-20b?
llm-speed has not measured gpt-oss-20b on RTX 4090 itself yet, so there is no llm-speed verdict. 2 community-submitted runs exist (fastest 242.6 tok/s at Q4_K_M on llama.cpp); community-submitted, not measured by llm-speed, so they give no verdict.
2 signed runs on llm-speed. Single-request decode speed in tokens per second; the short-chat workload unless noted. All 2 runs are community-submitted (signed on the submitter's own machine, not measured by llm-speed, and not used for llm-speed verdicts). 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 df18bee | 242.6 | chat-short | 1 | Community-submitted |
| MXFP4 | ollama | 0.31.1 | 141.7 | chat-long | 1 | Community-submitted |
llm-speed has not measured this pair itself, so there is no llm-speed verdict or fastest quant. The rows above are community-submitted, not measured by llm-speed.
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_t-a9v_ipdit | 2026-10-14 | chat-short | Q4_K_M | llama.cpp · commit df18bee | 166 | 256 | 242.6 | 593.8 | 280 | – | Community-submitted |
| r_t-a9v_ipdit | 2026-10-14 | chat-long | Q4_K_M | llama.cpp · commit df18bee | 3188 | 1024 | 231.2 | 7864.1 | 405 | – | Community-submitted |
| r_iu2sfa9ykvw | 2026-07-02 | chat-long | MXFP4 | ollama · 0.31.1 | 3188 | 1024 | 141.7 | 631.5 | 5048 | – | Community-submitted |
| r_t-a9v_ipdit | 2026-10-14 | agent-trace | Q4_K_M | llama.cpp · commit df18bee | 3287 | 840 | 236.1 | 13439.2 | 150 | – | Community-submitted |
| r_t-a9v_ipdit | 2026-10-14 | concurrent-decode | Q4_K_M | llama.cpp · commit df18bee | 1024 | 256 | 239.6 | – | – | – | Community-submitted |
Questions
Can an RTX 4090 run gpt-oss-20b?
llm-speed has not measured gpt-oss-20b on RTX 4090 itself yet, so there is no llm-speed verdict. 2 community-submitted runs exist (fastest 242.6 tok/s at Q4_K_M on llama.cpp); community-submitted, not measured by llm-speed, so they give no verdict.
What did community runs measure for gpt-oss-20b on an RTX 4090?
2 community-submitted runs (community-submitted, not measured by llm-speed): Q4_K_M on llama.cpp: 242.6 tok/s; MXFP4 on ollama: 141.7 tok/s. They are listed for reference and give no llm-speed verdict.