Indra
Verdict: Fast CPU-only OCR from Datalab that outputs reading-ordered markdown.
Use this if you need to turn scans or PDFs into markdown without a GPU.
Live X Buzz Score
X 34/100 ↗ Rising updated 15 Oct, 15:13 HKT
Daily X posts, last 7 days (oldest → today): 0 · 0 · 0 · 0 · 0 · 11 · 11. Score parts: volume 13.6/50, spike 13.5/25, engagement 6.8/25. Method
Why it's buzzing
@superdoccimo · 2026-10-14100万パラメータのOCRが、CPUだけで毎秒49ページ読む。GPUなしの文書処理パイプラインが一段安くなるかもしれない話 Marker（GitHub約4.0万スター）、Surya（約2.1万）、Chandra（約1.2万）で文書AIの定番OSSを出してきたDatalabが、10月13日に新しいOCRパッケージ「Indra
♥ 1 · ⟲ 1 · 💬 0 · 👁 582@OrgVio · 2026-10-13Datalab introduced Indra, a 1M-parameter OCR model that processes 49 pages per second on a 32-core CPU, delivering 11-17x speedup over PaddleOCR and Tesseract while achieving higher text extraction scores on OmniParseBen
♥ 0 · ⟲ 0 · 💬 2 · 👁 48@TechFollowrazzi · 2026-10-14🗞️ @soumithchintala followed @VikParuchuri @VikParuchuri released Indra OCR, built around a 1M-parameter recognition model. Datalab reports 49 pages per second on 32 CPU cores, with no GPU. https://t.co/SPbTlyPf7X http
♥ 4 · ⟲ 0 · 💬 0 · 👁 121Discussion
Comments are coming with sign-in.