Llama 3.2 1B Instruct
A lightweight instruction-tuned language model released by Meta in 2024 with approximately one billion parameters. It is designed for edge devices and supports quantized compression for local execution on consumer-grade hardware.
Timeline (last 90 days)
The tests used Transformers.js 4.2.0 to run three models on WebGPU with 4-bit weights: SmolLM2 360M Instruct, Qwen2.5 0.5B Instruct, and Llama 3.2 1B Instruct (q4f16).
With the full chat setup and example prompts, the three models produced 0 schema-conforming replies out of 72; after switching to greedy decoding, 25/72 conformed.
Touch Grass Bingo uses the Llama 3.2 1B Instruct open-weight model to generate 5x5 bingo cards for users.
When a card is generated for the first time, the browser downloads approximately 700 MB of model files and caches them locally on the device; all subsequent inference is performed locally on the device without going through a server.
Llama 3.2 1B Instruct is a lightweight instruction-tuned model released by Meta in 2024 with approximately 1 billion parameters, designed specifically for edge devices.
Version 1 required the model to return a JSON array containing 24 strings, but the 1B model on mobile phones repeatedly returned malformed JSON; after 3 retries it would fall back to displaying the built-in starter card.
All Facts (6)
The tests used Transformers.js 4.2.0 to run three models on WebGPU with 4-bit weights: SmolLM2 360M Instruct, Qwen2.5 0.5B Instruct, and Llama 3.2 1B Instruct (q4f16).
50%UnverifiedWith the full chat setup and example prompts, the three models produced 0 schema-conforming replies out of 72; after switching to greedy decoding, 25/72 conformed.
50%UnverifiedTouch Grass Bingo uses the Llama 3.2 1B Instruct open-weight model to generate 5x5 bingo cards for users.
50%UnverifiedWhen a card is generated for the first time, the browser downloads approximately 700 MB of model files and caches them locally on the device; all subsequent inference is performed locally on the device without going through a server.
50%UnverifiedLlama 3.2 1B Instruct is a lightweight instruction-tuned model released by Meta in 2024 with approximately 1 billion parameters, designed specifically for edge devices.
50%UnverifiedVersion 1 required the model to return a JSON array containing 24 strings, but the 1B model on mobile phones repeatedly returned malformed JSON; after 3 retries it would fall back to displaying the built-in starter card.
50%