Needle
An extremely lightweight edge-side AI model released by NVIDIA, designed to run locally on resource-constrained edge devices such as Raspberry Pi, smartwatches, and smart home appliances without requiring cloud infrastructure support.
Timeline (last 90 days)
Needle targets on-device scenarios, with model sizes ranging from 8 to 29 MB; it accomplishes tool calling, structured extraction, and text embedding at 2 bits per weight, and the inference engine is under 1 MB.
Needle outperforms models ten times its size on mobile tool calling on the project's own benchmarks, and returns an empty list rather than a wrong guess when uncertain.
Neural 3 (also known as Needle) is an on-device model launched by San Francisco startup Cactus Compute, incubated by YC; it requires no internet connection and can run on phones, watches, and small development boards.
NVIDIA's Needle model is a lightweight AI model that can run on-device without cloud support.
A project called Needle (6.7k stars) compresses a model down to just 14MB, enabling it to run directly on a smartwatch.
All Facts (4)
Needle targets on-device scenarios, with model sizes ranging from 8 to 29 MB; it accomplishes tool calling, structured extraction, and text embedding at 2 bits per weight, and the inference engine is under 1 MB.
50%UnverifiedNeedle outperforms models ten times its size on mobile tool calling on the project's own benchmarks, and returns an empty list rather than a wrong guess when uncertain.
50%UnverifiedNeural 3 (also known as Needle) is an on-device model launched by San Francisco startup Cactus Compute, incubated by YC; it requires no internet connection and can run on phones, watches, and small development boards.
50%UnverifiedNVIDIA's Needle model is a lightweight AI model that can run on-device without cloud support.
50%