Hacker News · AI·· 41 min agoAI score62
DocBlocks releases docbench: a 24-model local small-model benchmark for document tasks
Benchmarks for document tasks for local AI models
AI summary
The DocBlocks team has released docbench, which scores 24 local models ranging from 0.8B to 35B parameters across six categories of document tasks—review, writing, charts, slides, study aids, and transcript cleanup—using 102 test cases (142 outputs).
Source:Hacker News · AI · bendyline.com