Unverified50% confidenceEventExact time
In the September 22 update, the official adjusted the default precision of Qwen-Image GGUF from F16/BF16 to Q4_K_M.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/23/2026
First Seen
Valid until: 12/22/2026
Sources
Related Entities
Related Claims
VerifiedIn GGUF quantization naming, Q4_K_M denotes a 4-bit quantization, Medium-precision variant using the K-Quant grouped quantization method68% similarUnverifiedVia --quantization quark_mxfp4, SGLang dequantizes NVFP4 weights and requantizes them to MXFP4 at load time, achieving GSM8K accuracy recovery of 97.5%–100.2% on models such as MiniMax-M2.7 and GLM-5.1.67% similarUnverifiedQwen 2.5 7B使用4组GQA配置，推理时KV Cache体积相比MHA减少约75%66% similarUnverifiedThe 4-bit quantization of Qwen3 27B shows robust performance, with the gap from FP16 baseline on mainstream benchmark tasks remaining within an acceptable range.66% similarUnverifiedThe fine-tuned Qwen3.5-4B model improved accuracy on BU Bench V1 from 15% to 53%.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/942293API
curl https://kongchang.com/api/v1/knowledge/claims/942293MCP
get_claim(id=942293)