RTX 5090 for local LLMs
17 measured results from 6 builds, each linked to its source.
| Memory | 32 GB per card |
|---|---|
| Memory bandwidth | 1,792 GB/s (memory bandwidth as listed on llamabench) |
| Typical EU price | ≈€6,199 new (Asus GeForce RTX 5090 (lowest variant), billiger.de, 2026-10-16) |
Can it run? DeepSeek V4.1 Flash on RTX 5090 GLM-5.3 Flash on RTX 5090 Gemma 4 12B on RTX 5090 Gemma 4 26B-A4B on RTX 5090 Gemma 4 31B on RTX 5090 Ornith 1.5 122B on RTX 5090 Qwen3.5 9B on RTX 5090 Qwen3.6 27B on RTX 5090 Qwen3.6 35B-A3B on RTX 5090 Qwen3.8 27B on RTX 5090 Qwen3.8 Flash-Next on RTX 5090 gpt-oss 20B on RTX 5090
All builds with the RTX 5090
| Build | GPU(s) | Memory GB (total) | Results |
|---|---|---|---|
| Intel i9-13900K + RTX 5090 | RTX 5090 | 32 | 1 |
| Single RTX 5090 32GB | RTX 5090 | 32 | 1 |
| RTX 5090: 2026 models (llamabench community) | RTX 5090 | 32 | 7 |
| RTX 5090 (llm-speed signed runs, Oct 2026): gpt-oss 20B 324 tok/s, Qwen3.8 27B 75 tok/s | RTX 5090 | 32 | 3 |
| Ryzen 9 9950X3D + RTX 5090 + 128 GB DDR5: four big MoEs with experts offloaded to RAM (Qwen3.8 Flash-Next 78.8 tok/s) | RTX 5090 | 32 | 4 |
| RTX 5090 workstation: Qwen3.8 Flash-Next IQ3_S at 159.2 tok/s on Strata (expert offload) | RTX 5090 | 32 | 1 |
All measured results
(MTP, DFlash, DSpark and similar draft-and-verify methods give much higher tok/s on predictable text and are not comparable 1:1 with plain decode)
All RTX 5090 builds: see also all GPUs · best LLMs by memory size · download the dataset.
FAQ
How fast is the RTX 5090 for local LLMs?
The fastest plain-decode result we have is 324.2 tok/s on gpt-oss 20B (reported as Q4_K_M, llama.cpp 0.7.0-dev) in RTX 5090 (llm-speed signed runs, Oct 2026): gpt-oss 20B 324 tok/s, Qwen3.8 27B 75 tok/s. Speed depends heavily on the model, quant and engine, so compare like with like in the table. RigBase lists 17 measured results from 6 builds.
How much memory does the RTX 5090 have?
32 GB per card. Memory bandwidth: 1,792 GB/s (memory bandwidth as listed on llamabench).
Which models have been measured on the RTX 5090?
DeepSeek V4.1 Flash, GLM-5.3 Flash, Gemma 4 12B, Gemma 4 26B-A4B, Gemma 4 31B, Llama 3.2 3B, Ornith 1.5 122B, Qwen3.5 9B, Qwen3.6 27B, Qwen3.6 35B-A3B, Qwen3.8 27B, Qwen3.8 Flash-Next, gpt-oss 20B.
Where do these numbers come from?
Every row links to a build page with its original public source (benchmark logs, READMEs, forum posts). RigBase does not re-run the benchmarks. Results marked ⚡ use speculative decoding and are not comparable 1:1 with plain decode.
Referral link: if you sign up with Google and add $10, Runpod gives you and RigBase bonus credit (terms). It never affects our data.