Unverified50% confidenceSolutionExact time
The tester found that raising CFG from 1 to 4–4.5 yielded some improvement in output image quality.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/20/2026
First Seen
Valid until: 12/19/2026
Sources
Related Entities
Related Claims
UnverifiedThe tester's initial configuration was CFG 1, Steps 40, resolution 2MP, aspect ratio 4:3, sampler Euler, scheduler Simple, with the model and text encoder using int8 quantization and the VAE at bf16 precision.73% similarUnverifiedNVFP4量化方案相比BF16基线实现了1.5倍的推理加速，同时在top-1准确率上保留了92%到97%的水平67% similarUnverifiedIn mixed-precision training, each FP16 parameter occupies 2 bytes of GPU memory, and each FP32 parameter occupies 4 bytes of GPU memory.66% similarUnverifiedFor Qwen Image 2.1, it is recommended to start around CFG 4, paired with approximately 40 sampling steps.64% similarUnverified基于尾调用的字节码分发在某些基准测试中可带来 5-15% 的性能提升64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/937398API
curl https://kongchang.com/api/v1/knowledge/claims/937398MCP
get_claim(id=937398)