AI Briefing#ocr
The latest AI and developer news about #ocr, with the original source and a short summary.
The latest AI and developer news about #ocr, with the original source and a short summary.
The new benchmark uses 330+ error types to weight parsing errors by severity, correcting previous metrics that penalized formatting over factual errors.
Cactus Compute's Mirage model uses 51.3M parameters to perform multi-task visual analysis on mobile NPUs, enabling text-only LLMs to answer visual questions without fine-tuning.
The 1.3B parameter model uses a recursive self-improvement loop and pluggable layout analysis to convert PDFs into structured data.
The 2B parameter model achieves a 0.834 field-level F1 score, surpassing Claude Opus 5 by 1.8 points and GPT-5.6 Luna by 9.3 points.
The 689M parameter model supports 91 languages and processes documents at 4.5 pages/s on a single H100 GPU.
The new setting uses variable billing to handle complex documents, raising Value F1 from 77.7 to 94.3 on ExtractBench Pro.
The 1B OCR model runs locally on mobile devices while the 2B verifier model checks output accuracy against original images.
Baidu has released Unlimited-OCR 2, which uses cross-page context to parse long documents.
The working draft defines 140 rules for handling tables, forms, and diagrams to resolve inconsistencies between different parsers.
The new platform offers token-based billing with $5 in free credits for new accounts and supports synchronous parsing for documents up to 50 pages.
Billing for AI Search begins November 1, 2026, with a free tier offering 5M ingestion tokens and 1,000 semantic queries per month.
GravityOCR, based on GLM-OCR, increased decoding speed by 1.32x without adding parameters while maintaining an accuracy score of 95.16.
A 'home advantage' pattern was discovered where, among 13 document data extraction benchmarks, the manufacturer's models ranked in the top 3 in all 10 benchmarks they participated in, and took first place in 7 of them.
Inflearn is developing a 'Video Screen Translation' feature that recognizes text within video frames and overlays translated text.
The MCP tool 'Watch Skill' for Claude Code has been released, enabling it to analyze screen recordings to verify code changes.
Upstage AI won the Bronze Stevie for 'AI Company of the Year' and a Silver Stevie for CEO Kasey Roh at the 2026 American Business Awards.
A technique combining Diffusion draft generation and AR verification with the GLM-OCR model has been released, increasing decoding speed by 3.85x.
OpenAI's GPT-5.6 lineup demonstrated significant performance improvements in vision tasks such as object detection and counting.
The Sol model in OpenAI's GPT-5.6 lineup achieved overwhelming performance on vision benchmarks.
Mistral has released OCR 4.1, featuring paragraph-level bounding box extraction and structural block labeling capabilities.
Mistral AI has released OCR 4.1, featuring paragraph-level Bounding Box extraction and structural block labeling capabilities.
Mistral has released **Mistral OCR 4.1**, featuring paragraph-level bounding box extraction.
CohereLabs released a 2.4B vision-language model under the Apache 2.0 license.
SIE unifies serving of embeddings, reranking, OCR, and LLMs through a single OpenAI-compatible API.