RTX 5090 vs RTX 3090
MeasuredOn gpt-oss-20b (same quant and runtime family, chat-short), the RTX 5090 decoded at 324.2 tok/s and the RTX 3090 at 207.2 tok/s, each side's fastest matched run for that model.
Signed, community-submitted decode tok/s, prefill, and TTFT for RTX 5090 vs RTX 3090 — every number links to the run it came from.
How to read these results
On gpt-oss-20b, the RTX 5090 decoded at 324.2 tok/s and the RTX 3090 at 207.2 tok/s. Each row in the table is a like-for-like pair: the same model and size, quant and runtime family on the chat-short workload, using each side's fastest such run. Runs can still differ in host, runtime build and context. These are reference results, not a controlled speed or answer-quality ranking. Inspect both source runs before choosing hardware. 1 more model was measured on both but not like-for-like; they are listed separately in the "Not a like-for-like pair" table below and left out of the comparison.
Models that have data on both rigs
| Model | RTX 5090 decode | RTX 3090 decode | Δ | Source runs |
|---|---|---|---|---|
| gpt-oss-20b | 324.2tok/s chat-short · llama.cpp · Q4_K_M | 207.2tok/s chat-short · llama.cpp · Q4_K_M | +117.0 | r_dwp5fgdapmw · r_nakw1nq6clw |
| Qwen3-Coder-30B-A3B-Instruct | 304.9tok/s chat-short · llama.cpp · Q4_K_M | 199.7tok/s chat-short · llama.cpp · Q4_K_M | +105.2 | r_vsype4pm1ko · r_0dh1kzbv4h0 |
| Qwen2.5-Coder-14B-Instruct | 137.3tok/s chat-short · llama.cpp · Q4_K_M | 78.21tok/s chat-short · llama.cpp · Q4_K_M | +59.0 | r_xnr_k4emoyr · r_rkyf9wxzw9z |
| Qwen3.6-27B | 75.09tok/s chat-short · llama.cpp · Q4_K_M | 42.73tok/s chat-short · llama.cpp · Q4_K_M | +32.4 | r_pob-co51bd- · r_qm1xyvjmx_r |
| Qwen3.8-27B | 75.19tok/s chat-short · llama.cpp · Q4_K_M | 37.65tok/s chat-short · llama.cpp · Q4_K_M | +37.5 | r_qfrb2zmyk0k · r_ptct8o01o5l |
| Qwen3-32B | 69.83tok/s chat-short · llama.cpp · Q4_K_M | 38.45tok/s chat-short · llama.cpp · Q4_K_M | +31.4 | r_vhbqyt1f_xr · r_km62sybxga6 |
Not a like-for-like pair
These have data on both sides, but the model size, quant or runtime differs, so they are not a head-to-head and are left out of the comparison above.
| Model | RTX 5090 decode | RTX 3090 decode | Why it isn't matched | Runs |
|---|---|---|---|---|
| NVIDIA-Nemotron-3.5-Bolt-74B-A9B-24GB | 157.3tok/s llama.cpp · quant unreported | 87.75tok/s llama.cpp · quant unreported | quant unreported | r_ymx58emedxd · r_-sifd5otbpy |
See also: RTX 5090 benchmarks · RTX 3090 benchmarks · All hardware · All models · Methodology
Request the missing runs
No measured run yet for these pairs. Vote for the one you want; the most-requested runs go first. See all requests.
We never sell or share your email. Privacy