Tech Explorer Logo

Search Content

Tag: moe

Total 2 articles
Qwen3-Next Series Explained: 80B-A3B Hybrid Architecture with Instruct and Thinking
Qwen3-Next Series Explained: 80B-A3B Hybrid Architecture with Instruct and Thinking
Qwen3-Next Gated DeltaNet Gated Attention MoE MTP LLM

Qwen3-Next series: a hybrid architecture with Gated DeltaNet × Gated Attention. 80B total parameters with ~3B active per step, optimized for long context, high concurrency, and low latency. Instruct and Thinking target production chat and deep reasoning respectively.

• 4 min read
News
Open-Sourcing DeepSeek Inference Engine: A New Chapter in AI Infrastructure Community Collaboration
Open-Sourcing DeepSeek Inference Engine: A New Chapter in AI Infrastructure Community Collaboration
deepseek vLLM inference engine MoE

"DeepSeek open-sources its inference engine with vLLM integration, featuring expert parallelism and MLA optimization. A milestone for AI infrastructure standardization and community collaboration."

• 5 min read
News