AI1 min read
Jina AI launched jina-ocr-v1, a 3.4B parameter document parser with built-in speculative decoding.
Why it matters: Users can parse PDFs and scans faster on cheaper hardware like the NVIDIA L4.
By OpenSmartRoute editorial · written through the router by llm-onprem
From MarkTechPost - “Jina AI Releases jina-ocr-v1: A 3.4B MoE Document Parser With Built-In Speculative Decoding for Low-Budget GPUs”
Jina AI released jina-ocr-v1 as an end-to-end visual document parser.
It converts PDFs, scans, tables, charts, and invoices into clean Markdown in one pass. The model has 3.4 billion total parameters with about 570 million active per token. A speculative decoding head is included inside the checkpoint. Jina AI designed this system to run on low-budget GPUs such as the NVIDIA L4.
Keep reading
AI1 min read
Nvidia leaders Nader Khalil and Sydney Sykes will lead a session on open versus proprietary AI models. The event takes place at TechCrunch Disrupt 2026 in San Francisco.
AI1 min read
PrismML launched Bonsai 2, a 5.9 GB model based on Qwen3.8. It compresses the original by nine to ten times while keeping most performance.
Written through OpenSmartRoute
The router read the source excerpt, chose the model below among the deployment's executable targets and layered the editorial-writer skill on the prompt. The same request is one row in this deployment's public counters.
llm-onpremollama-qwen3-5-4b/qwen3.5:4beditorial-writerDrafted automatically from the source and reviewed automatically; it may contain errors. Refer to the source for the authoritative text. How routing works
Nunchux AI launched VC-Attention to speed up video diffusion models without retraining. It uses low-bit quantization and replaces the slow softmax stage.