OnDesk

What fits on the desk. Not in the cloud.

Pick a GPU, or open Model and search a repo. Fit numbers are ours, not Gemini. This page answers the question before you buy.

Catalog & repo links via Hugging Face Hub

8 GB vram · 272 GB/s

6–8 cores (Ryzen 5 / Core i5), PCIe 4.0 · 16–32 GB RAM

System RAM besides VRAM. Low end ≈ VRAM; high end ≈ 2× for OS and offload.

Speed
Whisper Large v31.6B3.84FP16perfect——
DeepSeek R1 Distill Qwen 1.5B1.5B3.76FP16perfectest. 25
Qwen2.5 1.5B Instruct1.5B3.76FP16perfectest. 25
Qwen2.5 Coder 1.5B1.5B3.76FP16perfectest. 25
Llama 3.2 1B Instruct1.2B3.17FP16perfectest. 30
TinyLlama 1.1B Chat1.1B2.8FP16perfectest. 34
Canary 1B1.0B2.71FP16perfect——
Gemma 3 1B Instruct1.0B2.63FP16perfectest. 36
Whisper Large v3 Turbo810M2.3FP16perfect——
Florence-2 Large770M2.83FP16perfectest. 34
Whisper Medium769M2.19FP16perfect——
Distil-Whisper Large v3760M2.19FP16perfect——
BGE M3570M2.79FP16perfect——
Qwen2.5 0.5B Instruct490M1.46FP16perfectest. 65
BGE Large EN v1.5340M1.83FP16perfect——
E5 Large v2340M1.83FP16perfect——
GTE Large340M1.14FP16perfect——
mxbai embed large v1340M1.14FP16perfect——
RF-DETR Large320M2.31FP16perfect——
SAM ViT-Large308M2.26FP16perfect——
Whisper Small244M1.14FP16perfect——
Florence-2 Base230M1.65FP16perfectest. 58
SAM 2 Hiera Large224M1.73FP16perfect——
Grounding DINO Tiny172M1.46FP16perfect——
RF-DETR Medium150M1.37FP16perfect——
Nomic Embed Text v1.5140M1.02FP16perfect——
BGE Base EN v1.5110M0.9FP16perfect——
E5 Base v2110M0.9FP16perfect——
Snowflake Arctic Embed M110M0.9FP16perfect——
SAM ViT-Base91M1.14FP16perfect——
SAM 2 Hiera Base+81M1.08FP16perfect——
RF-DETR Small80M1.08FP16perfect——

Showing 32 of 106.