Llama 3 8B Instruct Lite

Run It Yourself?

Llama 3 8B Instruct Lite ships open weights (8B parameters). Rough VRAM need — computed as parameters × bytes-per-weight × 1.2 overhead: ~20GB at FP16 or ~5GB with 4-bit quantization.

Cheapest live GPU servers clearing the 4-bit bar (5GB):