RNGD (pronounced “Renegade”) is Furiosa’s second-gen chip for high-performance, highly efficient inference in data centers with large language (LLM) and multimodal models. The chip went into production this summer and is now running advanced models like Meta’s Llama 3.1 70B at full speed using Furiosa’s RNGD software stack.

“Evaluating the product with customers’ real world deployments is just the first step toward achieving our goals with RNGD,” said Furiosa co-founder and CEO June Paik, “but we are racing forward quickly.” (Just look at June’s hair blowing in the wind in the photo below.)

Over the coming weeks and months, these customers will test RNGD’s performance in demanding real world applications, where latency, throughput, and energy consumption are all crucial considerations.

Initial feedback has been very positive, but we have more work to do to demonstrate the chip’s capabilities. Our APAC Tech Sales Director, Yungbum Jung (he’s the one without the flowy hair in the photo), and his team will work closely with customers to onboard them smoothly and collect their feedback.

RNGD is designed to deliver a unique combination of performance, power efficiency, and programmability all in a single product. We’re ramping up production now ahead of general availability in early 2025.

Sign up here to get updates on RNGD or reach out to us to learn more.