Liquid AI brings LEAP to AMD Ryzen laptops

Liquid AI announced on August 18, 2025, that its Liquid Edge AI Platform (LEAP) now natively supports AMD's Ryzen and Ryzen AI processors. The support covers the "Zen 5" generation with integrated Radeon graphics, and its purpose is to let developers deploy Liquid's LFM models on PCs instead of relying on cloud inference.

The PC integration is built on llama.cpp, which Liquid calls the de facto standard for inference on consumer hardware. Instead of writing native code for every combination of x86 and ARM chips, GPU vendor and operating system, the company is building on existing tools. It offers Python and Node.js bindings, and Liquid says the same code runs anywhere llama.cpp is supported.

Liquid benchmarked LFM2-1.2B in Q4_0 GGUF format with the public llama-bench executables, testing every pre-built backend and 4, 8 and 12 CPU threads. It concluded that Ryzen AI chips gave the best LEAP performance of the consumer hardware it tested. The integrated GPU handles prompt processing at speeds the company calls comparable to discrete GPUs, while the Zen 5 CPU handles token generation. The post does not give the throughput figures or name the competing hardware.

AMD CTO Mark Papermaster said the combination delivers "sub-100-millisecond responsiveness on-device." Liquid CTO Mathias Lechner described the update as "collapsing complexity for developers."

The SDK is available now, with documentation and example code. On the roadmap are LFM-2-VL vision-language support, more memory and power optimizations using GPUs and NPUs, and support for inference engines other than llama.cpp.