
Beam model offers frontier reasoning with efficient MoE architecture
The Beam model uses a sparse Mixture-of-Experts architecture to deliver high-level reasoning with 3-4x less inference compute than similar models.
The AI news that matters for people who make things — what happened, and how to put it to work in Grafy.

The Beam model uses a sparse Mixture-of-Experts architecture to deliver high-level reasoning with 3-4x less inference compute than similar models.

Grafy Research Lab reported on K2-Horizon, a suite of six open‑source language models ranging from 0.9B to a 375B‑A23B MoE.

Ant Ling has unveiled Ling-3.2-flash, a multimodal Mixture-of-Experts model with a 1M context window. Discover its benchmark performance and scale.

Thinking Machines Lab released Inkling 2 Small, a 594B MoE with 17B active parameters scoring 31 on the Artificial Analysis index. What it means for users.

Reflection's Beam is a 501B-parameter open model designed for long-context coding and agentic tasks.

A new C-based inference engine called Janas streams mixture-of-expert LLMs directly from an NVMe SSD.