EmbeddingGemma 2: an open, lightweight multimodal embedding model
deepmind.google·6d ago
AI Engineering
Lobsters·Aug 18·5 min read
Alibaba's TSMC-Built 5nm RISC-V Chip, XuanTie C950, Now Runs Qwen-3.8 27B Model Natively, Unlocking Massive Vertical Integration Tailwinds
TL;DR
Alibaba has developed the XuanTie C950, a RISC-V chip that natively supports its Qwen-3.8 27B AI model, enhancing its competitive edge in AI. This integration allows Alibaba to reduce reliance on external hardware and customize its AI solutions.
✦ Why It Matters
Consider exploring RISC-V architecture for your next AI hardware project to reduce costs and increase customization.
Key Takeaways
- Explore the XuanTie C950 for edge AI applications.
- Utilize the Qwen-3.8 model for efficient AI processing.
- Consider the benefits of RISC-V architecture for customization.
- Evaluate the chip's performance metrics for your workloads.
How It Works
The XuanTie C950 integrates matrix and vector acceleration engines directly into its architecture, allowing it to execute AI workloads without the need for GPUs. Its 16-stage pipeline and 8-instruction decode width enable high throughput and efficient processing of complex commands, making it particularly suited for running large language models natively.