Unverified50% confidenceFactExact time
Moondream forces each weight in the Parakeet encoder into one of three values—−1, 0, or +1 (ternary quantization), requiring approximately 1.58 bits of storage per weight.
1
Sources
50%
Confidence
Long-term
Relevance
10/6/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe model compresses the cache along three dimensions: 64 query heads share a single latent, every two adjacent tokens in the encoder are compressed into one entry, and only 4 out of 40 layers retain cache while the remaining 36 layers store nothing.67% similarUnverifiedBonsai uses ternary quantization, compressing each weight to one of three values—-1, 0, or +1—with a shared scaling factor per 128 weights, requiring approximately 1.7 bits per weight, roughly 9 times smaller than 16-bit full precision.66% similarUnverifiedTernary representation theoretically requires only about 1.58 bits per parameter, which can yield nearly a 10x storage compression compared to FP16.64% similarVerified4-bit quantization compresses each parameter to 0.5 bytes, theoretically compressing 140GB down to approximately 35–40GB.62% similarUnverifiedFor 240 million domain names, the memory footprint of a Succinct Trie can be compressed to only 2–4 bits per character, keeping the total within 1–2 GB.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/979388API
curl https://kongchang.com/api/v1/knowledge/claims/979388MCP
get_claim(id=979388)