Multimodal models and image generation: text-to-image, diffusion, and vision.
Fri, July 3Open Source
Interfaze open-sources diffusion-gemma-asr-small, a diffusion-based multilingual ASR modelInterfaze open-sources diffusion-gemma-asr-small, a multilingual ASR model that transcribes via diffusion instead of autoregression.
Fri, July 3Agent
Alibaba Page Agent: JavaScript GUI Agent for Web Control via Natural LanguageAlibaba's Page Agent reads the DOM directly via client-side JavaScript to execute natural language commands, no screenshots or multimodal models needed.
Wed, July 1Model/API
NVIDIA Releases Nemotron-Labs-TwoTower Diffusion Language ModelNVIDIA released an open-weight diffusion language model built on a frozen autoregressive backbone.
Wed, July 1Model/API
Google Launches Faster, Cheaper Image Generator Nano Banana 2 LiteGoogle updates its image generator to be faster, cheaper, and more useful for creators.
Tue, June 30Tools
Cost-Optimized Document Processing with Amazon Nova 2 Lite and ClaudeAWS shows how to combine Amazon Nova 2 Lite with Anthropic's Claude Sonnet 4.6 in a two-model pipeline on Amazon Bedrock to digitize scanned documents at low cost and large scale.
Mon, June 29Open Source
EverOS: Open-source Markdown-first, hybrid BM25+vector retrieval, self-evolving skills for AI agent memoryEverMind open-sourced EverOS, a local-first memory runtime that stores AI agent memory as plain Markdown, indexed via SQLite and LanceDB, with hybrid BM25+vector retrieval, multimodal ingestion, and self-evolving skills.
Sat, June 27Agent
Lyto: A Single AI Agent Across Browsers, Tools, and MessagesLyto launches a unified AI agent operating across browsers, tools, and messages.
Sat, June 27Model/API
ByteDance Diffusion LM iLLaDA Matches Qwen2.5ByteDance and Renmin University unveil 8B parameter diffusion language model iLLaDA, matching Qwen2.5 base performance.
Fri, June 26Tools
Cewsco: All-in-One AI Assistant Integrating Chat, Image, Voice, and Market DataCewsco is a versatile AI assistant combining chat, image generation, voice interaction, and market data.
Wed, June 24Research
DFlash Speculative Decoding: Parallel Token Blocks Boost Throughput Up to 15x on NVIDIA BlackwellUC San Diego's DFlash replaces autoregressive draft generation with a lightweight block diffusion model, generating entire token blocks in a single forward pass for accelerated speculative decoding.
Tue, June 23Research
Embedding the World: Multimodal AI for Searchable Aerial ImageryAn AWS blog describes building a large-scale searchable aerial imagery system using multimodal AI on Amazon Bedrock and OpenSearch Serverless, along with its evaluation.
Thu, June 18Model/API
Midjourney Launches First Ultrasound Full-Body Scanner HardwareMidjourney releases a full-body scanner based on ultrasound, pivoting from cat image generation to medical imaging.