Model Architecture

Mixture-of-Experts Mixture of Experts Transformer Models Sparse Mixture-of-Experts Multimodal Models Hybrid Models Parameter Efficiency Inference Optimization Dense Models Sparse Models

3 ARTICLES

yesterday

Nvidia Unveils Nemotron 3 Ultra, a 550B Open-Weight AI Model

Publishing weights and training recipes, Nvidia aims to cut inference costs and speed research by using a mixture-of-experts design with NVFP4-driven throughput gains.

9 ARTICLES

yesterday

Nvidia Unveils Cosmos 3, an Open World Model for Robots and Vehicles

8 ARTICLES

5d ago

Chinese AI Models Collapse Per‑Token Costs, Forcing Developers to Rearchitect How They Use LLMs

3 ARTICLES

3w ago

Google Releases Multi‑Token Prediction Drafters for Gemma 4 With Up to 3x Faster Inference

6 ARTICLES

last mo.

Gemma 4 Goes Local With Tool Calling on Phones and Apple Silicon

3 ARTICLES

last mo.

Google Releases Gemma 4, an Apache-Licensed Open AI Built for Local and Single-GPU Use

6 ARTICLES

2mo ago

Nvidia Launches Nemotron 3 Super, an Open 120B Model for Agentic, Long‑Context AI

12 ARTICLES

5mo ago

Moonshot AI Releases Open-Source Kimi K2 Thinking, a Tool-Using 'Thinking Agent'

8 ARTICLES

7mo ago

DeepSeek-AI Open-Sources 3B DeepSeek-OCR to Compress Visual Context for Long-Document OCR

3 ARTICLES

9mo ago

ByteDance’s Seed Team Open-Sources 36B-Parameter Seed-OSS Models With 512K Context

3 ARTICLES

10mo ago

Alibaba Deprecates Hybrid Mode in Qwen3, Launches Dedicated Instruct and Thinking Models

9 ARTICLES

10mo ago

Z.ai Unveils Open-Source GLM-4.5 Models That Undercut DeepSeek

7 ARTICLES

10mo ago

Alibaba Elevates Qwen3 Open-Source AI with Benchmark-Topping Upgrades

13 ARTICLES

last yr.

Alibaba Releases Qwen 3, Advancing Open-Source AI Competition

7 ARTICLES

last yr.

Meta Faces Backlash Over Use of Experimental AI Model for Benchmark Testing

20 ARTICLES

last yr.

Meta Launches Llama 4 AI Models, Introducing Scout and Maverick

7 ARTICLES

last yr.

DeepSeek V3 Challenges AI Giants With Open-Source Model and Efficiency Breakthroughs

Model Architecture

Nvidia Unveils Nemotron 3 Ultra, a 550B Open-Weight AI Model

Nvidia Unveils Cosmos 3, an Open World Model for Robots and Vehicles

Chinese AI Models Collapse Per‑Token Costs, Forcing Developers to Rearchitect How They Use LLMs

Google Releases Multi‑Token Prediction Drafters for Gemma 4 With Up to 3x Faster Inference

Gemma 4 Goes Local With Tool Calling on Phones and Apple Silicon

Google Releases Gemma 4, an Apache-Licensed Open AI Built for Local and Single-GPU Use

Nvidia Launches Nemotron 3 Super, an Open 120B Model for Agentic, Long‑Context AI

Nvidia Launches Nemotron 3 Open Models With Data and RL Toolkits

Mistral 3 Launches With Open-Weight Frontier Model and Edge Suite

Mistral AI Releases Open-Weight Mistral 3 Models Under Apache 2.0

Never miss stories about

Model Architecture

Moonshot AI Releases Open-Source Kimi K2 Thinking, a Tool-Using 'Thinking Agent'

DeepSeek-AI Open-Sources 3B DeepSeek-OCR to Compress Visual Context for Long-Document OCR

ByteDance’s Seed Team Open-Sources 36B-Parameter Seed-OSS Models With 512K Context

Alibaba Deprecates Hybrid Mode in Qwen3, Launches Dedicated Instruct and Thinking Models

Z.ai Unveils Open-Source GLM-4.5 Models That Undercut DeepSeek

Alibaba Elevates Qwen3 Open-Source AI with Benchmark-Topping Upgrades

Alibaba Releases Qwen 3, Advancing Open-Source AI Competition

Meta Faces Backlash Over Use of Experimental AI Model for Benchmark Testing

Meta Launches Llama 4 AI Models, Introducing Scout and Maverick

DeepSeek V3 Challenges AI Giants With Open-Source Model and Efficiency Breakthroughs

Never miss stories about

Model Architecture