#mixture-of-experts
Mixture-Of-Experts: 27 AI articles covering mixture-of-experts news, analysis, and research
Articles
Recursive Unveils Flint, a Laptop-Ready Coding Model Largelyβ9
Recursive releases Flint, a free coding model that runs locally on a laptop and was built largely by AI, an early test of recursive self-improvement.
MiniMax Launches M3.1 Flash: A 428B MoE Agentic Model at $0.15β8
MiniMax releases M3.1 Flash, a 428B MoE agentic model with 1M context at $0.15 per 1M input tokens, offering near-frontier performance at Flash-tier cost.
StepFun Releases Step 5 Preview as Open Weights: 600B MoE, 27Bβ9
StepFun releases Step 5 Preview open weights: a 600B MoE model with 27B active parameters, 1M-token context, and strong long-context reasoning for agentic workl...
NVIDIA Releases Nemotron 3.5 Super VL, a 120B Open Multimodalβ9
NVIDIA launches Nemotron 3.5 Super VL, a 120B open multimodal agent model activating 12B parameters per token, with a 1M context window for enterprise AI.
Meta Unveils Muse Spark Open: An Open-Weight Frontier Multimodalβ8
Meta unveils Muse Spark Open, an open-weight frontier multimodal mixture-of-experts model on Hugging Face, advancing efficient, accessible frontier AI for resea...
Fish Audio Releases Drama 3: A 35B Open-Weight MoE Text-toβ8
Fish Audio releases Drama 3, a 35B open-weight MoE text-to-speech model with 3B active parameters, running on a single 48GB GPU for scripted dialogue and dubbin...
Liquid AI Releases LFM2.5-24B-A2B: A 2B-Active MoE Built for Onβ8
Liquid AI releases LFM2.5-24B-A2B, a 24B MoE with 2B active parameters built for on-device agents, delivering fast local tool calling on laptops and PCs.
One Model Family, Two Gold-Level Results: Fine-Tuning NVIDIAβ9
NVIDIA Nemotron fine-tuning achieves gold-level results in both IOI competitive programming and IMO math olympiad, showing one foundation model family can maste...
Reflection AI Unveils Beam: A 501B Open-Weight MoE Model Withβ8
Reflection AI's Beam is a 501B-parameter open-weight MoE model with 23B active parameters, built for coding and agentic workloads with 3-4x lower inference comp...
Reflection Debuts Beam, an Open-Weight AI Model Taking Aim atβ9
Reflection AI launches Beam, a 501B-parameter open-weight reasoning model it claims matches top Chinese open models like GLM-5.2 at far lower inference compute ...