Skip to content
OpenSmartRoute

Blog

Research - news and analysis

Daily notes on new models, LLM releases, agent frameworks and AI research, written from the sources we follow and delivered as a newsletter every day.

Get the daily issue

Every new post of the day, in one email. Confirmation required.

AllModelsLLMsAgentsResearchAI
#models 156#llm 106#agents 96#ai 88#research 82#inference 39#news 28#evaluation 25#benchmark 22#training 16#automation 13#deployment 13#security 13#optimization 10#aws 9#open-weights 9#nlp 8#cybersecurity 7

Research1 min read

CARE-X: Radiology VLMs with Auxiliary Supervision

Microsoft Research introduced CARE-X, a new approach to radiology VLMs using auxiliary supervision, reward-aligned learning, and tool-augmented measurement for chest X-ray interpretation. This system aims to create clinically useful models with calibrated predictions and flexible reasoning.

From Microsoft Research

Research1 min read

Orchard: Open Framework for AI Agent Research

Microsoft Research released Orchard, an open-source framework designed to simplify the training and evaluation of AI agents. This framework focuses on reducing complexity and enabling strong performance from smaller models, facilitating research in scalable agentic AI.

From Microsoft Research

Research1 min read

Echoverse: Environments for Agent Training

Microsoft Research introduced Echoverse, a system that trains computer-use AI agents in deep, evolving environments. This approach addresses agent struggles with multi-step workflows by providing realistic and dynamic training scenarios.

From Microsoft Research

Research1 min read

EvoLib: LLMs Evolve Through Experience

Microsoft Research introduced EvoLib, a system that transforms deployment experiences into evolving knowledge for LLMs. This allows models to adapt and learn across tasks long after initial deployment, improving performance and reducing the need for retraining.

From Microsoft Research

Research1 min read

Our approach to bioresilience

Google DeepMind and Isomorphic Labs are sharing our joint approach to bioresilience and AI models.

From Google DeepMind blog

Research1 min read

Verifying Rust Cryptography Code with SymCrypt

Microsoft Research developed a method to verify cryptographic code written in Rust, focusing on SymCrypt. This approach aims to ensure code integrity while maintaining performance and adaptability during implementation and evolution.

From Microsoft Research

Research1 min read

Aurora 1.5 Enhancements for Weather and Earth Systems

Microsoft Research released Aurora 1.5, expanding the foundation model with 22 new variables, hourly resolution, and ensemble forecasting. This update improves its suitability for real-world weather, climate, and energy applications.

From Microsoft Research

Research1 min read

Flint: Visualization Language for AI Agents

Flint is an open-source visualization language enabling AI agents to generate expressive charts from concise specifications. It provides a middle ground between simple chart specifications and complex manual chart creation.

From Microsoft Research

Research1 min read

DeepMind and A24 Collaborate on Novel AI Research

Google DeepMind and A24 have initiated a research partnership focused on developing advanced AI agents. The collaboration aims to explore the use of large language models in creative workflows, specifically for scriptwriting.

From Google DeepMind blog

Research1 min read

Nano Banana 2 Lite and Gemini Omni Flash Released

Google DeepMind has released Nano Banana 2 Lite, a smaller language model, and Gemini Omni Flash, a new flash storage solution optimized for AI workloads. These tools are designed to enable efficient and cost-effective AI development and deployment.

From Google DeepMind blog

Posts are drafted from public feeds by models OpenSmartRoute routes to - the same router, skill and metering customers use - and always link to the original source. Corrections: support.

How this blog is made

Every post is a routed request

Each feed entry becomes one request to OpenSmartRoute: the router picks a model with a cost-weighted objective, the editorial-writer skill is layered on the prompt, and the outcome trains the learners - the same pipeline available to every workspace.

Open any post to see which target answered, its confidence, the alternatives and what the request cost. Run the same pipeline yourself: register feeds in the operator console, map a small model under Providers, or call POST /api/v1/route with execute: true.