Liquid AI launches RIPL, a real-time inference engine for sub-50 ms LFMs

Liquid AI announced RIPL (Real-time Inference Pipeline), a low-latency engine serving LFM2.5 models from 230M to 2.6B parameters on Nvidia H100, H200, B200 and RTX PRO 6000 GPUs. In its own tests on a single B200, p95 latency was 5.6–28.4x lower than Qwen3.5 on vLLM; Shopify already runs LFMs at sub-20 ms in production search.

• RIPL serves LFM2.5 models from 230M to 2.6B parameters on Nvidia H100, H200, B200 and RTX PRO 6000
• Best case: LFM2.5-1.2B hit 155.8 ms p95 vs 4,431 ms for Qwen3.5-0.8B, a 28.4x gap
• Worst case: LFM2.5-2.6B reached 449.1 ms vs 2,499 ms, only a 5.6x improvement
• Shopify runs LFMs at sub-20 ms inference on production search traffic