从语言到影像，从开源到 API。读懂模型能力、比较真实成本，
为你的下一个想法，找到合适的智能引擎。
This is the instruct model of Dream 7B, which is an open diffusion large language model with top-tier performance. More details about the model and usage can be found in the blog and github bellow: - **Blog:** https://hkunlp.github.io/blog/2025/dream/ - **Github:** https://github.com/HKUNLP/Dream
译文排队中 · 当前显示来源原文
基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
基础资料已收录 · 缺失资料后续补充
基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
基础资料已收录 · 缺失资料后续补充
基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
基础资料已收录 · 缺失资料后续补充
基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
基础资料已收录 · 缺失资料后续补充
基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
基础资料已收录 · 缺失资料后续补充
**MinerU-Popo** is a lightweight and universal framework for POst-Processing OCR outputs, bridging the gap between page-level OCR parsing and document-level semantic structure. It construct document tree structure based on with a 4B post-processing model performing four subtasks: table truncation analysis, text truncation analysis, title hierarchy analysis, and image-text association analysis. We handle the challenges of cross-page geometric discontinuity, redundant document parsing and scalability to long documents via: **Basic OCR** | **Before** | **After** :---:|:---:|:---:| MinerU | 53.7 | **90.6** | MonkeyOCR | 48.9 | **87.4** | Dolphin | 60.4 | **83.5** | PaddleOCR | 59.3 | **82.6** | GLM-OCR | 53.5 | **81.8** | **Model** | **TEDS** | **Doc/s** :---:|:---:|:---:| MinerU-Popo | **90.6** | **0.37** | Qwen3-VL-2B | 21.2 | 0.22 | Qwen3-VL-4B | 56.5 | 0.20 | Qwen3-VL-8B | 65.9 | 0.16 | Qwen3-VL-32B | 78.0 | 0.04 |
译文排队中 · 当前显示来源原文
An abliterated version of Google's Gemma 3 12B IT created using Heretic. This model has reduced refusals while maintaining model quality, making it suitable as an uncensored text encoder for video generation models like LTX-2. Check out Gemma 3 12b it Heretic v2 which has vision capabilities, NVFP4 support and decensored with Heretic v1.2.0. It'd be better to use v2. **Note:** GGUF support in ComfyUI for Gemma text encoders is experimental. See PR #402 for status. The GGUFs work with llama.cpp directly.
译文排队中 · 当前显示来源原文
An abliterated version of Google's Gemma 3 12B IT created using Heretic v1.3.0. This model has reduced refusals while maintaining model quality, making it suitable as an uncensored text encoder for video generation models like LTX-2. Available in five quantization formats for ComfyUI - FP8, INT8 (ConvRot row-wise), INT4 (W4A4 ConvRot), NVFP4, and MXFP8 - all produced with SVD-guided learned rounding for maximum fidelity, covering everything from Ada GPUs to Blackwell. You can see the docker, scripts and configurations used to make these files on Heretic Docker Github.
译文排队中 · 当前显示来源原文
An abliterated version of Qwen 3 4B created using Heretic v1.2.0. This model has reduced refusals while maintaining model quality, making it suitable as an uncensored text encoder for image generation models like Z-Image and FLUX.2 Klein 4B. Available in five ComfyUI-native quantized formats (FP8, INT8, INT4, NVFP4, MXFP8), all produced with SVD-guided learned rounding for maximum fidelity. > **Quality:** All quantized variants use SVD-guided learned rounding (AdaRound via convert_to_quant), which optimizes each weight's rounding direction to minimize output reconstruction error — noticeably higher fidelity than naive round-to-nearest quantization. All variants load natively in ComfyUI 0.30.0+ (no plugins) via the `comfy_quant` metadata embedded in each file.
译文排队中 · 当前显示来源原文
基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
基础资料已收录 · 缺失资料后续补充
基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
An abliterated version of Qwen's Qwen3-8B created using Heretic v1.2.0 (git master). This model has reduced refusals while maintaining model quality, making it suitable as an uncensored text encoder for image generation models like Klein 9B. Available in five quantization formats for ComfyUI: FP8, INT8 (ConvRot row-wise), INT4 (W4A4 ConvRot), NVFP4, and MXFP8, all produced with SVD-guided learned rounding for maximum fidelity. You can see the docker, scripts and configurations used to make these files on Heretic Docker Github.
译文排队中 · 当前显示来源原文
基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
基础资料已收录 · 缺失资料后续补充
GGUF quantisations of the Heretic abliteration of Qwen3-VL-4B-Instruct. The full forensic report comparing all candidate variants sits in the main repo. Several Heretic trials were run against Qwen3-VL-4B and all of them reach 100% HarmBench ASR, up from 30.8% on the base. This build was picked because, with safety tied, it wins on the tie-breakers: > **Vision note:** these GGUF builds target the text path. For Qwen3-VL vision support, use the bf16 / transformers repo or a ComfyUI checkpoint.
译文排队中 · 当前显示来源原文
An abliterated version of Qwen's Qwen3-VL-8B-Instruct created using Heretic v1.3.0. This model has reduced refusals while maintaining model quality. You can see the docker, scripts and configurations used to make these files on Heretic Docker Github. ComfyUI converted text encoders are provided in various formats, along with the original bf16 safetensors. This is a vision-language model (supports both text and image inputs). It does **not** support thinking/reasoning tokens.
译文排队中 · 当前显示来源原文
基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
基础资料已收录 · 缺失资料后续补充
Lucid is a family of AI language models developed and used by DreamGen with a **focus on role-play and story-writing capabilities**. Note that reasoning is mostly intended for assistant tasks (knowledge, math, coding, etc.) and much less for writing. The smaller Lucid models ( `writer character Jon` -> `writer character Amy` -> `user` -> `assistant` -> `writer narration` -> ... - You should use `writer character Jon Doe` when sending a message for a particular character (in this case Jon Doe), **including the user character**. - You should use `user` when providing some instructions (OOC) to the model. The model understands structured scenario information, consisting of plot, style, setting, characters, etc. (this is not an exhaustive list, and all sections are optional).
译文排队中 · 当前显示来源原文
基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
基础资料已收录 · 缺失资料后续补充
基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
基础资料已收录 · 缺失资料后续补充
Add **anime** to your prompt to make your gens look more anime. Add **photo** to your prompt to make your gens look more photorealistic and have better anatomy. This model was trained on 768x768px images, so use 768x768px, 704x832px, 832x704px, etc. Higher resolution or non-square aspect ratios may produce artifacts. **1girl**, **girl**, etc. give a bit different results, feel free to experiment and see which one you like more! Use this model as well as Dreamlike Diffusion 1.0 and Dreamlike Photoreal 2.0 for free on dreamlike.art!
译文排队中 · 当前显示来源原文
基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
You can add **photo** to your prompt to make your gens look more photorealistic. Non-square aspect ratios work better for some prompts. If you want a portrait photo, try using a vertical aspect ratio. If you want a landscape photo, try using a horizontal aspect ratio. This model was trained on 768x768px images, so use 768x768px, 640x896px, 896x640px, etc. It also works pretty good with higher resolutions such as 768x1024px or 1024x768px. This model can be used just like any other Stable Diffusion model. For more information, please have a look at the Stable Diffusion Pipeline.
译文排队中 · 当前显示来源原文
可以调整厂商、分类或输入输出条件；新模型正持续收录。
调整月度 Token 用量，直观比较不同模型的预估 API 开支。
使用已采集且单位可比的 USD / 百万 Token 渠道报价。报价为采集时快照；不计缓存优惠、阶梯价、推理额外输出与税费，请以来源的当前报价为准。
先区分输入、输出和缓存价格，再使用同一任务和用量估算总费用。图像、音频和视频可能按张、时长或分辨率计费，不能直接与 Token 单价比较。
不一定。必须逐一核对具体模型版本的许可证、商用限制及使用政策。自己部署还需要考虑显存、推理资源和运维费用。
榜单取决于测试任务、版本与评测方法。建议用实际业务样本比较准确性、时延、稳定性和成本，再选择模型。
