Tag: open source
NVIDIA’s Nemotron 3.5 Lightning routes agent grunt work to small models
NVIDIA released Nemotron 3.5 Lightning, a 30B open model for the busy work inside agent runs, plus a…
Meta’s Glimmer release puts desktop agents in reach
The Apache-licensed 30B model brings agentic AI to local machines.
Microsoft open-sources Orchard to train agents in real harnesses
Microsoft's Orchard framework trains and evaluates coding and browsing agents inside the same harnesses they will ship in.
Tencent opens Team Memory hub with no fix for wrong facts
Tencent's Team Memory gives AI agents a shared knowledge hub, but critics say wrong facts have no correction…
AgentRadio lets coding agents share notes without pausing
AgentRadio's async messaging lets coding agents share discoveries mid-task and nearly double their accuracy.
DeepMind’s cyclone model beats forecasters by a day, then goes open
WeatherNext forecasts cyclones with a day more warning, and Google is releasing the weights.
Mistral opens 3B Shieldstral model for adaptive content safety
Mistral's Apache-licensed 3B classifier takes safety policies as plain-language questions at inference time, policing text and images with…
AMD’s open 16B MoE model tops fully open rivals on benchmarks
AMD's fully open mixture-of-experts model leads open rivals while activating just 2.8B parameters.
NVIDIA’s Molt framework slims agentic RL to readable code
NVIDIA's NeMo team opens Molt, a compact PyTorch-native framework that makes agentic RL research easier to read and…
Inkling Small matches flagship skills at a quarter of the size
Thinking Machines shipped a quarter-size Inkling model that nears the flagship's benchmark scores.
MCP revamp drops sessions to scale AI agent tool calls
The biggest Model Context Protocol revision since launch removes sessions and adds extensions, authorization, and a deprecation policy.
Amodei clears the air on open-weights stance
Anthropic CEO refutes claims the company wants to ban open-weights AI models, calling for targeted policies instead.