Liquid AI has open-sourced Pipette, a new benchmarking suite designed to evaluate the performance of on-device AI models. Developed in partnership with Artificial Analysis, Pipette measures models by considering the entire deployment system, including quantization, runtime, and hardware, rather than just the model in isolation. The platform aims to provide reproducible results for developers and companies deploying AI on edge devices, with an initial dataset covering over 1,000 configurations across 30+ models and various hardware. AI
IMPACT Enables more accurate performance evaluation of AI models on edge devices, crucial for consumer electronics and embedded systems.
RANK_REASON The release is an open-source tool for benchmarking, not a new frontier model release or core research paper.
- Artificial Analysis
- Galaxy S26 Ultra
- Granite-4.0-350M
- Granite-4.0-H-350M
- iPhone 17 Pro
- Liquid AI
- llama.cpp
- MacBook Pro
- Qwen3.5-4B
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →