Unleashing the Power of Tenstorrent Galaxy Blackhole: Revolutionizing AI Video Generation and LLM Inference
Tenstorrent introduces the Galaxy Blackhole, a cutting-edge solution for general-purpose AI that offers exceptional performance in AI video generation and Large Language Models (LLMs) prefill and decode tasks. Unlike other solutions that require combining separate accelerators, Tenstorrent's Networked AI integrates compute, memory, and networking into a unified system optimized for real-world AI workloads.
The Tenstorrent Galaxy Blackhole stands out for its industry-leading performance across various AI workloads, excelling in video generation, LLM inference for prefill and decode, and supporting a wide range of model architectures. Experience the power of Tenstorrent Galaxy Blackhole at the launch event, TT-Deploy, on May 1st at 1:30pm PT through a live stream.
For AI video generation, the Tenstorrent Galaxy offers a significant speed boost, being 10 times faster than leading GPU systems. Collaborating with Prodia, the Galaxy Blackhole supercluster achieved a remarkable feat by generating a 720p, 81-frame video in just 2.4 seconds, showcasing unparalleled performance in video model processing.
In Blitz Mode, the Tenstorrent Galaxy excels in Large Language Model (LLM) inference tasks, delivering over 350 tokens per second and a sub-4-second time-to-first-token on Deepseek-R1-0528 671B. This outperforms leading GPU systems and enables high-margin AI use cases such as agentic workflows, real-time systems, and long-context reasoning.
Tenstorrent Galaxy sets new performance benchmarks in the AI industry, offering unmatched speed and efficiency for AI video generation and LLM inference tasks. Experience the power of Tenstorrent Galaxy Blackhole for yourself and witness the future of AI computing.