Sakana AI's technology has been adopted in "Evidence Finder," an evidence search tool for physicians from Iris Inc. (@ShoOkiyama). In response to physicians' questions, it searches the literature and answers while explicitly citing its sources. Sakana AI's technology handles the comparison and integration of literature and the generation of answers. For details, see here: https://sakana.ai/namazu-aillis/ 🐟
Leapfrog, a small team doing high-volume AI visual and production work for fashion and brand clients, is building a "one brain" system that makes company knowledge and client context searchable through natural-language agents. The starter stack described is OpenClaw in a sandbox, a GitHub repository, Obsidian on the local machine, and Telegram as the access point. The system's research structure had roughly 1,200 files at the time of the talk.
Our borderless Lakehouse lets Gemini query AWS and Azure with no variable egress fees, read directly from Salesforce Data 360, SAP, ServiceNow, and Workday without copying data, and federate open Apache Iceberg tables across Databricks Unity, Snowflake Horizon, and AWS Glue → https://goo.gle/3TT9y4X
3️⃣ Skills give Gemini the instructions, and tools give it the connection. But grounding gives it the context. Gemini grounded in your enterprise data delivers accurate, high-impact results.
Markdown is all you need. (Mostly.) A parser can get every word on the page right and still lose which column a number belongs to. Then your model has to guess. Markdown keeps headings, lists, and tables intact, and it stays readable when you're debugging a bad answer. For tables with merged headers, we switch to HTML. Read our breakdown on why it's our default output for parsing below! ⬇️
Happening now: Join our webinar to discuss the future of search and retrieval, including Embed 5, Parse 5, and our new retrieval methodology, RCP-nDCG https://x.com/i/broadcasts/1DxLdZyjNXvxm
Microsoft's Foundry blog guide advises keeping existing Azure Document Intelligence workloads that meet production requirements. It recommends evaluating Azure Content Understanding for high-variation, unstructured, reasoning, RAG, or multimodal document scenarios, and for new cloud OCR or layout workloads.
Perplexity is open-sourcing pplx-embed-v2-late, multi-vector embedding models for text and images in one shared space, in 9B and 0.6B sizes. The 9B model can index multimodal data, the 0.6B model can run queries on device, and PDF pages can be searched without OCR. The author reports 92.4% on MADQA and 64% on BrowseComp+, with weights available on Hugging Face.
On MADQA, 500 questions over 800 PDFs, it reaches 92.4%, the top result among retrievers.
In agentic search, a GPT-OSS-120B agent using the 9B model answers 64.0% of BrowseComp+ questions correctly, 4.9 points ahead of the next ColBERT model, with fewer searches than any baseline.
On Q2D-Web, about 70,000 production-derived search queries over 190M web documents, the 9B and 0.6B models reach 74.8% and 73.6% Recall@1000, against 69.3% for nemotron-embed-8b.
Both sizes are distilled token by token from one 18B teacher, so they share one embedding space. A corpus indexed with the 9B model can be searched with 0.6B queries. That lifts ViDoRe v3 from 62.3% to 63.5% with no added query cost.
The models also embed images and rendered pages directly, so PDFs, slides, and scans are searched without OCR, keeping tables, figures, and layout that text extraction drops.
Dense embedding models compress each document into one vector, which loses detail as pages get longer or more visual. pplx-embed-v2-late keeps a 128-dimensional vector per token and scores with MaxSim, so each query token is matched to its closest token in the document.
We're releasing pplx-embed-v2-late, two late-interaction embedding models that retrieve text, images, and pages with a shared embedding space for cross-model querying. Both models achieve frontier performance and are publicly available on Hugging Face. https://www.perplexity.ai/hub/blog/multimodal-embeddings-beyond-a-single-vector
Qlik built Qlik Answers, a natural-language assistant that returns sourced answers from knowledge bases, analytics apps, glossaries, and documents, using Amazon Bedrock for model access. The system routes each question through specialist agents and retrieval on Amazon OpenSearch Service, with Amazon Bedrock Guardrails applied to every request and response. Qlik serves more than 40,000 customers across regions, using Amazon SageMaker AI as an in-Region fallback when models are not yet available on Bedrock.
Google DeepMind released EmbeddingGemma 2, an open 740M-parameter embedding model that maps text, code, images, video, and audio into one 768-dimensional space. Text-only use needs a 270M-parameter footprint, about 191MB active RAM when quantized on a Pixel 11 Pro, while loading all modalities takes about 567MB. The reported MTEB Code NDCG@10 score is 78.68, about 14% above the first generation, and MTEB Multilingual v2 is 61.36, roughly flat.
Google DeepMind has released EmbeddingGemma 2, an open 740 million parameter model that maps text, images, audio, and video into one embedding space. It is built on the Gemma 4 architecture under an Apache 2.0 license and supports an 8K token context window. The company reports a code benchmark gain from 68.76 to 78.68 on MTEB Code and says the model can run on-device with about 567MB of active RAM for the full multimodal version on a Google Pixel 11 Pro.
AIWhy it matters: The release shows how a 740M-parameter embedding model can cover text, code, images, audio, and video on local hardware, with memory and storage figures to compare against other on-device options.
When paired with Gemma 4, EmbeddingGemma 2 powers efficient, on-device retrieval augmented generation (RAG) with a lower memory footprint. EmbeddingGemma 2 retrieves local files, and Gemma 4 reasons over them for grounded answers with complete privacy.
Google's new multimodal embedding setup processes image, audio, and video entirely offline with zero server calls. Its modular design lets developers drop unused vision and audio components to save memory, and flexible dimension sizes cut local database storage by up to 6x. It can also pair with Gemma 4 to build RAG pipelines with minimal memory and processing requirements.
Google DeepMind released EmbeddingGemma 2, a 740M-parameter embedding model, under an Apache 2.0 license. The post says it is competitive across benchmarks and outperforms some specialist models more than twice its size, and that developers can use it for multimodal search or pair it with Gemma 4 for on-device RAG. Weights are available on Hugging Face and Kaggle.
Google introduces EmbeddingGemma 2, its first open, natively multimodal embedding model, covering text, code, image, video, and audio tasks. It has a 740M parameter form factor, is positioned for offline, privacy-first RAG when paired with Gemma 4, and the post claims it outperforms some specialist models more than twice its size. Weights are available now on Hugging Face.
Dell is adding agentic AI capabilities to its AI Data Platform, including a Unified Semantic Layer with a searchable glossary and an Enterprise Knowledge Graph built with Nvidia's Auto-Ontology open source library. The features are designed to give agents shared context, reducing repeated token generation and compute costs. The platform's layers include the Data Orchestration Engine, Data Engines, and Storage Engines such as PowerScale, ObjectScale, and the Lightning File System.
Google released EmbeddingGemma 2, an open embedding model under the Apache 2.0 license that maps text, code, images, video, and audio into a shared 768-dimensional space. Developers can load a 270M-parameter text and code setup, or add vision and audio encoders up to a 740M-parameter full multimodal model. Matryoshka truncation to 256 or 128 dimensions reduces vector storage, with the guide noting quality losses on image, video, and speech retrieval at lower dimensions.
AIWhy it matters: The guide gives concrete encoder sizes and dimension-storage tradeoffs, showing how to choose a configuration for text, code, image, video, and audio retrieval.
A survey of 300 data, AI, and technology executives found only 34% of organizations' agentic AI projects reach production, with legacy systems, security concerns, and missing knowledge context as main obstacles. Production leaders, who advance 61% of projects beyond pilot, show stronger semantic knowledge capabilities. Most firms plan to invest in retrieval pipelines, AI-ready APIs, retrieval-augmented generation, and knowledge graphs.
Sovereign: Control where and how your models of choice run, with the ability to prototype documents, on-prem and air-gapped deployments, and shared knowledge for the whole org.
Jev can search 🔍 through a long document just like you would! 🕵️
Cloudflare introduced a Web Search API through AI Gateway, partnering with Ceramic.ai, Exa, and Linkup to give agents fresh web results instead of guessed URLs. Requests appear in AI Gateway logs and draw from AI Gateway credits, with partners committing to Cloudflare's Verified bots crawling standards and including source links in results. Partners at list API pricing without markup are available via a REST endpoint or a Workers binding, with native server tools planned.
Cloudflare's AI Search is now generally available, adding native image embeddings for visual retrieval, OCR for scanned PDFs, and a 10 MiB file limit up from 4 MiB. Billing begins November 1, 2026, charged per ingested token, stored GB-month, and query, with a free monthly allotment on all Workers plans.
On ConTEB, the preview has the highest average nDCG@10 of the models tested, though not on every task. It also beats voyage-context-4 on chunk retrieval while using 8x less storage per vector: 1 KB (1024 dims, int8) vs 8 KB (2048, float32).
pplx-embed-v2-context-9b-preview leads on Answer and Evidence retrieval at every cutoff. Document retrieval is closer, with our context v1 4B slightly ahead at Document@3 and Document@5.
context-bench is a benchmark for context-aware retrieval, created and privately held by @turbopuffer. Its queries, documents and capabilities are inspired by conversations with turbopuffer customers. It has 2,099 queries and 38,894 documents. We submitted for blind evaluation.
We overcome gold-chunk supervision limits by distilling relevance from our query-aware context compression model. It scores every document token against the query. Aggregated into chunk-level targets, those scores train the embedder to retrieve answer and supporting chunks.
Retrieval systems often split long documents into chunks, but this strips away the surrounding context. Contextual embedding models fix this by encoding the whole document once and pooling chunk vectors afterward. They are usually trained on one gold chunk per query.
We built a new way to train contextual embedding models, which encode each chunk of a document with the whole document in view. pplx-embed-v2-context-9b-preview sets a new state of the art on ConTEB and @turbopuffer's new, privately held context-bench. https://www.perplexity.ai/hub/blog/contextual-embedding-beyond-the-gold-passage
Kinda crazy new models from @Nils_Reimers and team. Extremely scalable, SOTA accuracy, and like always fully privately deployable and accessible on Model Vault!
Microsoft's Azure AI team argues that better models do not eliminate the need for a dedicated content extraction layer, since agents need trustworthy, structured, and auditable inputs. The post notes that building extraction directly on an LLM quickly demands chunking, layout parsing, grounding, normalization, and evaluation infrastructure. Microsoft positions Azure Document Intelligence and Azure Content Understanding in Foundry Tools as managed options for that layer.
More users shouldn't have to mean slower AI. In @Signal_65 RAG testing hosted via @TensorWave, AMD Instinct MI355X delivered ~42% lower p99 latency and roughly 2x the concurrent-user headroom before breaching the evaluated SLA: https://bit.ly/45hAvl3
LlamaIndex argues that frontier vision-language models still fail on real forms such as W-2s, 1040s, W-9s, and scanned W-4s, because forms require detecting every field, preserving section hierarchy, linking values to their exact boxes, and reading handwriting and checkmarks. The company's blog post details these failure modes and presents a custom cookbook for LlamaParse as a cheaper way to handle such forms.
GitHub's latest podcast episode examines five common AI hot takes, including whether developers must still read AI-generated code. It argues review effort should match risk, and that Skills and MCP solve different problems. It also says retrieval-augmented generation (RAG) remains useful and works alongside agents, skills, and MCP.