LlamaIndex Releases OCRia-1.5, Setting New SOTA on Tables, Charts and Forms Parsing
llama_index · x · 2026-11-19
LlamaIndex introduced OCRia-1.5, more accurate than OCRia-1 (4B) on complex tables, charts and forms, with more precise bounding boxes. Two sizes: flagship OCRia-1.5-4B powers LlamaParse's Agentic tier; new OCRia-1.5-2B has half the parameters and still beats the old model on all six dimensions of their ParseBench-2 document parsing benchmark. Both are live in LlamaParse today, setting state of the art on tables, charts and formatting on ParseBench-2.
Related event: LlamaIndex Releases OCRia-1.5 with Better Table and Chart Parsing(4 posts)→
More from Models
- RadixArk trains DaisyRose-27B in 18 days on 144 H200s, doubling Terminal-Bench score — BanghuaZ · 2026-11-19
- GPT-6.2 Astra tops Code Arena WebDev categories at 42% lower cost than predecessor — arena · 2026-11-19
- LLNL benchmark: only 2 of 9 AI models stay accurate when the simulator is secretly wrong — alex_verem · 2026-11-19
- Pichai demos Gemini 3: napkin sketches become websites, images become board games — heyneighbor · 2026-11-19
- OpenRouter token traffic: Chinese open-source models sweep top 5, no Western OSS in top 10 — PAstynome · 2026-11-19
- Code Arena launches WebDev leaderboard as GPT-6.2 Astra takes #1 in Consumer Product — arena · 2026-11-19