Unverified50% confidenceFactExact time
The Coder version retains 192 experts per layer (half pruned), with a transformer weight size of 72.1 GB, compared with 138.2 GB for the unpruned IQ2_XXS build.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
11/2/2026
First Seen
Valid until: 1/31/2027
Sources
Related Entities
Related Claims
UnverifiedIQ2_XS can compress a model to close to 2 bits/weight, enabling a 70B model that would otherwise require 48 GB of VRAM to fit into a 24 GB consumer-grade GPU72% similarUnverifiedDeepInfer's wide expert parallelism spreads 384 experts across 128 chips, with each chip holding only about 7 experts, reducing memory traffic by nearly 88%, and a single GPU can output more than 26,000 words per second.68% similarVerifiedMemory requirements are at least 37.6 GB for Q2_0, at least 39.2 GB for IQ2_XS, and at least 47 GB for IQ3_XXS, with the vision encoder occupying an additional 0.91 GB.68% similarUnverifiedCompressing DeepSeek Coder to 2-bit loses 67% of passing solutions, whereas the 4-bit version sacrifices only 2% for a 70% reduction in size.67% similarUnverified双编码器通常将语义压缩为768维向量，该数字来源于BERT-base的隐藏层大小，工业部署中常压缩到128或256维66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1127276API
curl https://kongchang.com/api/v1/knowledge/claims/1127276MCP
get_claim(id=1127276)