Liquid AI Launches New RIPL Inference Engine
The engine processes foundation models in under 50 milliseconds to support real-time tasks like search and payments.
Updated on Oct. 22, 2026 in Artificial Intelligence
Live Poll
Is faster AI inference more important than other performance metrics for your daily digital services?
Liquid AI has released its new Liquid RIPL inference engine, which serves foundation models directly within a live request path. The technology is designed to enable sub-50 ms response times for applications that require immediate model outputs.
Why it matters
Traditional LLM serving often struggles with the latency requirements of real-time operations like payment authorization and search ranking. By optimizing the full request-to-response path, this engine aims to make generative AI viable for high-speed industrial applications.
The engine operates on NVIDIA H100, H200, B200, and RTX PRO 6000 GPUs by executing hybrid architectures of convolution and GQA blocks. LFM2.5 models range from 230 million to 2.6 billion parameters, with structures varying from 16 to 30 layers.
The players
Liquid AI
An artificial intelligence company that develops foundation models and inference technologies.
Shopify
A multinational e-commerce company that provides a platform for online stores and retail point-of-sale systems.
NVIDIA
A technology company known for designing graphics processing units that power AI and high-performance computing workloads.
The details
Liquid AI claims the engine achieves a latency gap up to 28.4 times lower than current benchmarks on vLLM. Shopify has already deployed the technology for production search tasks, reporting inference speeds below 20 ms.
Timeline
Liquid AI released the Liquid RIPL inference engine on October 22, 2026.
The Tech Race
This development highlights an ongoing arms race to optimize inference speeds as AI models move from chat interfaces to critical backend infrastructure. By outperforming established benchmarks like the vLLM serving framework, Liquid AI positions itself against legacy model serving methods.
Users may experience faster search results and more responsive checkout processes on major e-commerce platforms like Shopify. These technical improvements aim to reduce wait times during digital transactions without compromising model capability.
The takeaway
Real-time AI integration requires significant architectural changes beyond simply shrinking model sizes. Organizations looking to adopt generative AI for latency-sensitive tasks must prioritize inference engine efficiency to maintain a seamless user experience.
Further reading
For more developments in this field, visit the Artificial Intelligence section.
Live Poll
Is faster AI inference more important than other performance metrics for your daily digital services?
