Ollama on the NVIDIA RTX 5080
On the NVIDIA RTX 5080, Ollama runs gpt-oss 21B · 3.6B active at 4-bit at a median 178.3 tokens per second for generation, from 1 plain run (16K context).
Speed by model on the NVIDIA RTX 5080
Plain runs only: one device, one request, no speculative decoding, the whole model in memory. Each row is one model size at one bit class, with the other engines' plain runs of the same model on this card beside it. Context length and build still differ between runs, and each side shows its own.
| Model | Ollama | Other engines |
|---|---|---|
| gpt-oss 21B · 3.6B active 4-bit | 178.3 t/sone report, 16K context | no other engine reported |
| Llama 3.2 3B 4-bit | 298.6 t/sone report, 16K context | no other engine reported |
| Qwen2.5 14B 4-bit | 93.4 t/sone report, 16K context | no other engine reported |
| Qwen2.5 14B 4-bit | 94.6 t/sone report, 16K context | no other engine reported |
| Qwen2.5 7B 4-bit | 176.6 t/sone report, 16K context | no other engine reported |
| Qwen2.5 7B 4-bit | 171.4 t/sone report, 16K context | no other engine reported |
8 Ollama reports on the NVIDIA RTX 5080
Newest first. A run on several cards, with speculative decoding, with part of the model in system RAM or with batched requests says so, since each describes a different setup.
- Model not identified · Q4_K_M34.9 t/s
- Qwen3-Coder 30B · 3B active · Q4_K_M51.7 t/s(part in system RAM)
- Qwen2.5 14B · Q4_K_M93.4 t/s
- Qwen2.5 7B · Q4_K_M171.4 t/s
- Qwen2.5 7B · Q4_K_M176.6 t/s
- Llama 3.2 3B · Q4_K_M298.6 t/s
- Qwen2.5 14B · Q4_K_M94.6 t/s
- gpt-oss 21B · 3.6B active · MXFP4178.3 t/s
Frequently asked
How fast is Ollama on the NVIDIA RTX 5080?
On the NVIDIA RTX 5080, Ollama runs gpt-oss 21B · 3.6B active at 4-bit at a median 178.3 tokens per second for generation, from 1 plain run (16K context).