#mixture-of-experts-moe

[ follow ]
#rubin-platform #nvidia #nemotron-3 #open-models #agentic-intelligence #llm-optimization #model-architecture
Python
fromPyImageSearch
2 days ago

Building and Training a Kimi-K2 Model Using DeepSeek-V3 Components - PyImageSearch

Agentic intelligence enables LLMs to perceive, plan, reason, and act through interaction, and Kimi-K2 delivers strong benchmark and leaderboard performance with architectural innovations.
Artificial intelligence
fromTechzine Global
4 months ago

Nvidia Blackwell successor Rubin releases in 2026: significant performance boost

Rubin is a six-chip AI infrastructure platform delivering up to 10× lower cost-per-token and faster training, available via major cloud providers in H2 2026.
#nvidia
fromZDNET
4 months ago
Artificial intelligence

Why Nvidia's new Rubin platform could change the future of AI computing forever

fromZDNET
4 months ago
Artificial intelligence

Why Nvidia's new Rubin platform could change the future of AI computing forever

more#nvidia
Artificial intelligence
fromComputerworld
4 months ago

Nvidia bets on open infrastructure for the agentic AI era with Nemotron 3

Nemotron 3 enables enterprises to build domain-specific AI agents without creating foundation models by providing open models, training data, and reinforcement-learning libraries.
Artificial intelligence
fromTheregister
5 months ago

Baidu teases next-gen AI training, inference accelerators

Baidu unveiled M100 (inference) and M300 (training) chips and will deploy Tianchi256/512 clusters to lower inference costs and support larger MoE models.
Artificial intelligence
fromZDNET
6 months ago

A new Chinese AI model claims to outperform GPT-5 and Sonnet 4.5 - and it's free

Moonshot's open-source Kimi K2 Thinking claims superior reasoning performance to GPT-5 and Sonnet 4.5 on several benchmarks while offering tool use and MoE architecture.
fromTheregister
6 months ago

How Perplexity optimized 1T parameter AI models for AWS EFA

AI search provider Perplexity's research wing has developed a new set of software optimizations that allows for trillion parameter or large models to run efficiently across older, cheaper hardware using a variety of existing network technologies, including Amazon's proprietary Elastic Fabric Adapter. These innovations, detailed in a paper published this week and released on GitHub for further scrutiny, present a novel approach to addressing one of the biggest challenges in serving large-scale mixture of experts models (MoE) at scale: memory and network latency.
Artificial intelligence
Artificial intelligence
fromLogRocket Blog
8 months ago

Agentic AI for 5x less: Why Kimi K2 is a frontend game-changer - LogRocket Blog

Kimi K2 is an open-source, agentic mixture-of-experts model that activates 32B of 1T parameters to execute tasks efficiently and rapidly.
[ Load more ]