chiprook
← AI
AIOctober 15, 2026, 06:37

Datalab releases OmniParseBench, an open OCR benchmark with 16,288 tests

Datalab has launched OmniParseBench, an open OCR benchmark for document parsers with 16,288 pass/fail tests across 2,937 pages from 2,343 documents in 92 languages. Gemini 3.8 Flash leads at 92.9, followed by Datalab accurate at 92.7. The dataset and Apache 2.0 code are public, and scoring runs offline without API keys.

Datalab releases OmniParseBench, an open OCR benchmark with 16,288 tests
#Datalab#Gemini
Read next
AI

Benchmark: Gemini 2.5 Flash fails false-premise test

AI

Datalab launches OmniExtractBench to fix bias in extraction benchmarks

AI

Google releases open Gemma 4 model family

AI

Gemini 4 Argon beats Claude Opus 5.5 in 14 of 18 benchmarks