According to Artificial Analysis, Ling-3.0-tiny sits on the mobile intelligence–speed Pareto frontier: 59 at 16K and 5.7s on iPhone 17 Pro. It also ranks first in the 64K intelligence evaluation with a score of 66.
Bringing stronger intelligence to smaller devices.
Thanks to @ArtificialAnlys for evaluating intelligence and real-device inference on the same phone-ready quantized builds.
Explore the full benchmark and methodology:
https://t.co/vHGpiekGRk
quantization — Shrinking a model by storing its numbers less precisely — like rounding — so it runs faster and fits on smaller hardware, at a small quality cost.
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
inference — Running a trained model to get answers — the phase where AI is actually used, as opposed to trained.
Why it matters
Ling-3.0-tiny sits on the measured intelligence-versus-speed frontier for phone-class inferenceRunning a trained model to get answers — the phase where AI is actually used, as opposed to trained.Full definition →, giving on-device builders a concrete quantizationShrinking a model by storing its numbers less precisely — like rounding — so it runs faster and fits on smaller hardware, at a small quality cost.Full definition → option with published latency on real hardware.
