Ollama v0.35.1-rc0 Adds Explicit Model Capability Declarations
Ollama released v0.35.1-rc0, introducing CAPABILITY declarations to Modelfiles and an additive capabilities field to create requests . The release preserves declarations across GGUF and Safetensors creation, inheritance, and Modelfile expor…
Nvidia Kumo Tabular Targets Tabular Prediction Accuracy-Efficiency Gains
Nvidia released Kumo Tabular, a model the company describes as setting a new accuracy-efficiency frontier for tabular prediction . Details on specific benchmark results, architecture, and availability were published on Hugging Face .…
Ollama v0.35.0 Adds Decision Models via /v1/systemone Endpoint
Ollama released v0.35.0, introducing support for decision models through a new /v1/systemone endpoint based on TypeSafe AI's Jev API . Decision models return choices, probabilities, and scores instead of text, targeting tasks such as ticket…
CoreWeave Open-Sources Agent Reinforcement Trainer for LLM Agents
CoreWeave released Agent Reinforcement Trainer (ART), an open-source framework for training large language model agents using reinforcement learning . ART includes support for rollouts, rewards, Group Relative Policy Optimization (GRPO), an…
ALL BRIEFS
RANKED · 5 THIS DAYOAuth credential theft flaw in MCP Python SDK forces patch across both release lines
A vulnerability in the official Model Context Protocol Python SDK allowed a malicious MCP server to steal OAuth credentials from any client application connecting over HTTP, Cycode reported and the SDK's maintainers confirmed in a security…