Unverified50% confidenceReleaseExact time
Conway Research released the model card for Saluki 27B, using IQ2 quantization to compress the 27B parameter model to approximately 7.89GB.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/10/2026
First Seen
Valid until: 1/8/2027
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 GPU76% similarVerified一个原本需要60GB显存的30B参数模型，经过量化后能在16GB甚至8GB显存的消费级显卡上运行71% similarVerified32B参数规模的模型在量化后通常可以压缩到约16-20GB显存占用，一张NVIDIA RTX 4090（24GB显存）就能运行68% similarUnverified动态GGUF的UD-IQ2_XXS档位将原始BF16模型体积从54.7GB压缩到约9GB，体积压缩率83.5%，准确率保留82.5%67% similarUnverified量化后70亿参数模型的显存需求从约28GB压缩至约4-6GB，普通消费级显卡如RTX 3060/4060即可流畅运行Llama 3、Qwen2.5、Mistral等70亿参数级别模型66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/994994API
curl https://kongchang.com/api/v1/knowledge/claims/994994MCP
get_claim(id=994994)