KK.
Back to work

Local VLM
Caption Benchmark

I wanted to see how well vision-language models could caption images on my laptop, and what they cost in time and memory.

4 models / 120 images / 480 captions / October 2026GitHub
Image Captioning on a Laptop
Open PDF Download
Page 1 of 11 of Image Captioning on a Laptop: Comparing the Quality and Efficiency of Local AI Models
Page 2 of 11 of Image Captioning on a Laptop: Comparing the Quality and Efficiency of Local AI Models
Page 3 of 11 of Image Captioning on a Laptop: Comparing the Quality and Efficiency of Local AI Models
Page 4 of 11 of Image Captioning on a Laptop: Comparing the Quality and Efficiency of Local AI Models
Page 5 of 11 of Image Captioning on a Laptop: Comparing the Quality and Efficiency of Local AI Models
Page 6 of 11 of Image Captioning on a Laptop: Comparing the Quality and Efficiency of Local AI Models
Page 7 of 11 of Image Captioning on a Laptop: Comparing the Quality and Efficiency of Local AI Models
Page 8 of 11 of Image Captioning on a Laptop: Comparing the Quality and Efficiency of Local AI Models
Page 9 of 11 of Image Captioning on a Laptop: Comparing the Quality and Efficiency of Local AI Models
Page 10 of 11 of Image Captioning on a Laptop: Comparing the Quality and Efficiency of Local AI Models
Page 11 of 11 of Image Captioning on a Laptop: Comparing the Quality and Efficiency of Local AI Models
Page 1 of 11Independent student research ยท October 2026
Next project LocalPilot