Together AI Announces Kernel Research Team and Blackwell GPU Optimization
Action Required
Together AI's kernel optimization capabilities can dramatically reduce inference latency and improve the performance of AI-native applications, leading to better user experiences and reduced operational costs.
AI Impact Summary
Together AI is highlighting its kernel research team's capabilities, specifically their ability to optimize AI models for NVIDIA Blackwell GPUs. The team's approach, rooted in academic research and hardware-software co-design, focuses on closing the performance gap between theoretical AI advancements and practical production deployments. This is demonstrated through projects like the Together Megakernel, which achieved significant latency reductions in voice agent applications, showcasing the team's ability to translate research into tangible business value for customers.
Affected Systems
- Date
- Date not specified
- Change type
- capability
- Severity
- high