Home › Models › Llama 3.1 8B Instruct › NVIDIA RTX 4090 Laptop
Llama 3.1 8B Instruct on the NVIDIA RTX 4090 Laptop
Llama 3.1 8B Instruct runs at about 49.2 tokens/s on the NVIDIA RTX 4090 Laptop (measured, Q4_K_M, llamafile, LocalScore).
Q4_K_M · llamafile
same run
Will it fit? Q4_K_M fits
| GGUF file, Q4_K_M | 4.9 GB | Hugging Face file size |
| Runtime + KV-cache allowance | 1.0 GB | estimate about an 8K-token context |
| Memory needed | 5.9 GB | |
| Usable on the NVIDIA RTX 4090 Laptop | 15.5 GB | 16 GB minus 0.5 GB for the driver |
Check another quant or memory size in the Will it run calculator
Every run
| Quant | Backend | Generate tok/s | Prompt tok/s | Test | Build / by / notes | Source |
|---|---|---|---|---|---|---|
| Q4_K_M | llamafile | 49.2 | 2,458 | LocalScore avg | LocalScore rank 43/431 | LocalScore (localscore.ai) |
tg128 = generating 128 tokens; pp512 = reading a 512-token prompt (llama-bench). LocalScore averages several prompt and output lengths, so its numbers run lower than tg128. "decode, llama-server" = single-stream generation speed measured against a running llama-server (no speculative decoding).
No card yet?
Try it on a rented GPU (RunPod)Affiliate link: if you sign up, RigBench may earn a referral credit. It never affects rankings or numbers.
All results for the NVIDIA RTX 4090 Laptop · Llama 3.1 8B Instruct on other hardware
Embed this result
HTML
<a href="https://rigbench.pages.dev/model/llama-3-1-8b-instruct/on/nvidia-rtx-4090-laptop/"><img src="https://rigbench.pages.dev/badge/llama-3-1-8b-instruct--nvidia-rtx-4090-laptop.svg" alt="Llama 3.1 8B Instruct · NVIDIA RTX 4090 Laptop: 49.2 tok/s (Q4_K_M, measured)"></a>
Markdown (GitHub README, forums)
[](https://rigbench.pages.dev/model/llama-3-1-8b-instruct/on/nvidia-rtx-4090-laptop/)