Introducing GLiNER2.5-Decide, a new classification encoder model fine-tuned for constrained decision making.

Open weight models are gaining adoption across the AI industry, and American research labs have an opportunity to catalyze innovation through openness and collaboration with labs around the world.

Companies are deploying open weight models in production. We examine early deployment strategies, evaluation practices, and system design.

A recap of the best open-weight models of summer 2026 and which model to reach for depending on the task.

From a PhD project to 45 million downloads: how GLiNER became GLiNER2.5, and why encoders still beat LLMs at extraction.

A new boundary-prediction architecture replaces span enumeration, adding joint entity-relation extraction, constrained classification, and unlimited span length.

Open-weight models now rival the frontier, closed models. Compare Kimi K3, DeepSeek V4, GLM 5.3, Qwen 3.8 and more, with pricing, benchmarks, and how to switch.

Two specialized open weight models for regulated industries, fine-tuned on NVIDIA Nemotron 3.5 Lightning by the Fastino Fine-Tuning Agent.

Learnings from fine-tuning Nemotron 3.5 Lightning with an autonomous agent.