Technology ❯ Artificial Intelligence ❯ Model Architecture

Mixture of Experts

Performance Metrics Multimodal Models Performance Optimization Efficiency Parameter Efficiency Contextual Memory Inference Efficiency Sparse Models Gemma 4 Kimi 3 Resource Optimization Parameters Inference Cost Context Length Scalability

Xiaomi Open‑Sources Two Agent Platforms: SoloEngine and MiMo Code

The releases signal a push to make long‑running, multi‑agent AI work cheaper for teams outside the developer community.

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

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