{"slug":"tiiuae-falcon-perception","kind":"claim","status":"testing","tier":"claim","title":"Falcon-OCR (270M): accurate full-page OCR from a tiny model?","category":"data-rag/doc-parsers","topic":"vision","repos":[{"repo":"tiiuae/Falcon-Perception","stars":786,"license":"Apache-2.0"}],"claims":[{"claim":"A 270M early-fusion OCR model, trained from scratch on OCR with no VLM pretraining, small model with amazing performance","source":"https://x.com/dahou_yasser/status/2105762692980170889"}],"plan":"Render 3 images with known text (clean print, small print, a table), run Falcon-OCR's end-to-end mode on the box GPU, and score character accuracy and word recall against the known text.","who_its_for":"People doing OCR who want a tiny local model instead of a big VLM.","explainer":"A 270M-parameter OCR model from TII that claims to punch far above its size. Rocky feeds it images whose exact text is known and measures how accurately it reads them.","measure":["Character accuracy vs known text","Word recall","Speed per image"],"notes_so_far":["Method: pinned c457916 (v1.0.0), model auto-downloaded from Hugging Face, GPU job. Rocky builds the images so the ground truth is known.","GPU test queued; results land here when it runs."],"sneak_peek":["First run on 3 images: clean text and small text both came back with 100% character accuracy. The table came back as HTML with all words recalled, but the scorer measured raw characters, so the table score was wrong.","A rerun with the 5-page OCR packet, scored by table cells and repeated 3 times, is queued."],"charts":[],"gallery":[],"results":{"jobs":[{"id":25,"title":"Falcon-OCR 270M: OCR 3 known images (print/small/table), score char-accuracy vs ground truth (GPU)","kind":"smoke","status":"done","started":"2026-10-18T05:14:07Z","finished":"2026-10-18T05:46:35Z","wall_seconds":1948,"install_friction":null,"artifacts":[]},{"id":33,"title":"Falcon-OCR claim: OCR packet 5 pages x3 repeats, table cell recall, install friction","kind":"claim","status":"failed","started":"2026-10-18T19:58:03Z","finished":"2026-10-18T19:58:05Z","wall_seconds":2,"install_friction":{"method":"README: clone + pip install -e .[torch]","seconds":0.0,"ok":false,"manual_fixes":0,"commands_run":2},"artifacts":[]}],"smoke_only":true,"ingested":"2026-10-18T20:31:32Z"},"last_test_finished":"2026-10-18T19:58:05Z","results_in":true}