Unverified50% confidenceFactExact time
Searching for Qwen Image 2.1 in Draw Things shows the original teacher model and the Turbo accelerated model, each offered in five quantization precisions; the version without a suffix is the 8-bit version, and the one with Exact is the FP16 version.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/21/2026
First Seen
Valid until: 1/19/2027
Sources
Related Entities
Related Claims
UnverifiedLess than a month after Qwen Image 2.1 was released, the official team launched an 8-step Turbo version.68% similarUnverifiedQwen-Image-2.1-Turbo is an accelerated checkpoint released 19 days after the base version Qwen-Image-2.1.67% similarUnverified第一代Qwen起步上下文窗口为2048，通过NTK插值外推到8192，训练使用Flash Attention、AdamW优化器和BF16混合精度67% similarUnverifiedTurbo and the base version use the same 7B visual generation architecture and are loaded with the same QwenImage21Pipeline.67% similarUnverifiedQwen3.8 27B officially provides a BFloat16 full-precision version and an FP8 version; Unsloth has released a 4-bit version with NVFP4 quantization.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1083023API
curl https://kongchang.com/api/v1/knowledge/claims/1083023MCP
get_claim(id=1083023)