/
#55
PrismML
Caltech-founded edge LLM startup. Builds the Bonsai model family, open-source LLMs with fully 1-bit weights across embeddings, attention, and MLP layers, no higher-precision escape hatches. Bonsai 8B fits in 1GB versus 16GB for FP16, runs on consumer CPUs and NPUs, and benchmarks competitively against Llama 3 8B.
Categories
Subcategories
IMAGE GENERATION
Links