Historical source metadata, not copies of the files. A checksum does not establish current availability.
model.safetensorseaad037756f1afd7cb847ff4b7c23db02ec56936bb30806903fb57d2b0b1588d
tokenizer.jsonbe75606093db2094d7cd20f3c2f385c212750648bd6ea4fb2bf507a6a4c55506
extras/ImageEditServer.pySHA-256 not recorded
extras/NVFP4TextEncoder.pySHA-256 not recorded
extras/ImageGenServer_cpu.pySHA-256 not recorded
extras/OmniImageEditServer.pySHA-256 not recorded
extras/QwenBackend.pySHA-256 not recorded
extras/ImageGenServer_new.pySHA-256 not recorded
extras/ImageGenClient.pySHA-256 not recorded
chat_template.jinjaSHA-256 not recorded
extras/ZImageTurboBackend.pySHA-256 not recorded
extras/KontextBackend.pySHA-256 not recorded
extras/ImageGenServer.pySHA-256 not recorded
config.jsonSHA-256 not recorded
extras/compress_mllm.pySHA-256 not recorded
extras/QwenImageBackend.pySHA-256 not recorded
extras/Flux2Backend.pySHA-256 not recorded
extras/GlmBackend.pySHA-256 not recorded
.gitattributesSHA-256 not recorded
extras/imagegen_zimage_turbo.shSHA-256 not recorded
tokenizer_config.jsonSHA-256 not recorded
README.mdSHA-256 not recorded
extras/imagegen_zimage_turbo_int4.shSHA-256 not recorded
recipe.yamlSHA-256 not recorded
generation_config.jsonSHA-256 not recorded
This is Z-Image-Turbo text encoder quantized to NVFP4, with working inference scripts that emulate OpenAI image generation viewpoint enclosed in extras. Scripts use vLLM for actual encoder inference. In combination with nunchaku-ai/nunchaku-z-image-turbo, the model comfortably fits on a 16GB NVIDIA GeForce RTX 5060 Ti GPU, allowing very fast image generation with no sequential offload needed. I also uncensored encoder with Heretic to hopefully improve prompt adherence.
Model discussions
Share what you know. Help others. Build trust.
Loading comments...
Keep it about the model. No spam, scams or personal attacks. Links are not allowed in comments or replies. Comment policy