RAG

Mistral OCR 4 adds layout-aware extraction with bounding boxes, block typing, and inline confidence across 170 languages. Use it through the API or self-hosted deployments when document pipelines need structure, citations, redaction, and chunking.

Turbopuffer said it crossed a $100M run-rate while staying profitable on less than $1M raised, and said Cursor moved production search onto the stack with a 95% cost reduction. The milestone matters because AI products increasingly compete on retrieval quality and cost, not just model output.

Firecrawl added a Highlights mode to /scrape that returns matching text, code, or tables for a query instead of full-page payloads. The release matters because the company benchmarked the feature on 10,000 URLs against Exa Highlights and aims it at lower-token agent retrieval.

Firecrawl introduced a /scrape mode that answers a question directly from a URL instead of returning chunks for a separate retrieval loop. It targets docs and pricing pages, and teams should use it when they want grounded answers with lower token usage.

Google expanded Gemini API File Search to index text and images together, add custom metadata filtering, and return page-level citations. RAG builders can use it for tighter retrieval control and more auditable answers.

IBM released 97M and 311M multilingual Granite Embedding R2 models under Apache 2.0, replacing XLM-RoBERTa with ModernBERT and extending context length from 512 to 32,768 tokens. The 311M model posts a +11.8 gain on MMTEB retrieval and ships with ONNX, OpenVINO, vLLM, and GGUF support.

BidirLM released a 2.5B multilingual encoder that embeds text, images, and audio into one shared 2048-dimensional space and works directly with Sentence Transformers. It tops several open-data embedding leaderboards and can run locally on GPU.

LightOn open-sourced DenseOn and LateOn plus the training pipeline behind them, including 1.4 billion query-document pairs and decontaminated BEIR results. Teams can use the small open retrieval models and reproduced data mixtures instead of opaque closed-data baselines.

OpenRouter added Firecrawl as a search provider, letting models ground responses in scraped full web pages instead of snippet-only search. The launch folds crawling into the existing plugin settings flow and includes a capped free plan on the Firecrawl side.

Sentence Transformers v5.4 adds one encode API for text, image, audio, and video, plus multimodal reranking and a modular CrossEncoder stack. It also flattens Flash Attention 2 inputs for text workloads, reducing padding waste and VRAM use.

LongTracer open-sourced local STS+NLI claim checks, while qi published a private search engine with a Claude Code plugin and LM Studio users shared MCP search configs for Qwen. Use these stacks to ground retrieval and verify answers without a second judge model.

Keep added an in-app feed reader so saved links can be read directly inside its bookmark store for agent workflows. Use it to turn bookmarks, RSS feeds, and markdown exports into reusable context instead of scattered tabs.

Google launched Gemini Embedding 2 in preview, unifying multiple modalities and 100+ languages in one embedding space with flexible output dimensions. Try it to simplify cross-modal RAG and search pipelines, but compare it with late-interaction systems before committing.

Google put Gemini Embedding 2 into public preview with one vector space for text, images, video, audio, and PDFs, plus 3072, 1536, and 768 output sizes. Use it to replace multi-model retrieval pipelines with one API for RAG and cross-media search.