Unverified50% confidenceFactExact time
The official algorithm for "1 bit per parameter" is: 72.12 GB divided by the original 551.9B transformer parameters gives 1.05 bits, rounded to 1; the removed experts are counted as zero bits.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
11/2/2026
First Seen
Valid until: 1/31/2027
Sources
Related Entities
Related Claims
UnverifiedThe follow-up paper "The Era of 1-bit LLMs (BitNet b1.58)" found that pure 1-bit requires very large scale to match full-precision quality, so it introduced a zero state, with weights becoming -1, 0 or +1.70% similarUnverifiedUnder 2-bit quantization, each weight has only about 4 discrete values.70% similarUnverified三值化权重每个权重仅携带约1.58 bits信息量（log₂3≈1.585），相比FP32的32 bits信息密度下降约20倍69% similarUnverifiedThe 280.1B parameters retained by Coder average 2.06 bits, no single weight is stored at less than 1 bit, and the pruned-away parameters lower the overall average by being counted as zero bits.69% similarUnverifiedBitNet 1.58b quantization uses ternary weights {-1, 0, +1} in PCEI's 24-layer network, reducing parameter memory by about 4 times compared with FP16 precision, and at inference time integer addition can replace floating-point multiplication.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1127277API
curl https://kongchang.com/api/v1/knowledge/claims/1127277MCP
get_claim(id=1127277)