This is an OpenVINO INT8 quantized version of empero-ai/Qwen3.8-9B-Distill, generated using optimum-intel. For detailed specifications, evaluation scores, and architecture info, please refer to the original repository: empero-ai/Qwen3.8-9B-Distill.
SAVRN Model Hub
Open-Weight Models
An open-weight model is an AI model whose trained weights are published for anyone to download. The weights are what the model learned in training. With a copy of them you can run the model on hardware you control and train it further on your own data.
Open weights are not the same as open source. Many publishers release the weights without the training data or code, and the license sets what you may do with the model. This library puts each model's full card, architecture, files, license and published evaluations on one page.
12,579Models
3,583Datasets
291Papers
7,578Publishers
12,850Sourced relationships
Updated 2026-10-23 · How the library is built
12,579 models, sorted by most downloaded.
Qwen Image 2.1 edits don't stay where you put them. Ask for a watercolor, a comic or an anime version and the picture comes back a few percent taller, sometimes nudged sideways, different for every seed: a waistband 40 px lower, a face that no longer lines up with the original. On a local edit it also repaints things you didn't ask about (hair strands, signs, texture). This LoRA makes the edit on the original's frame: same prompt, same edit, nothing moves. No trigger word. Load it and write your edit instruction as usual. Also on The dashed line marks where a feature sits in the original; the arrow is how far it moved. Same prompt and seed in every column, rendered in ComfyUI with the…
Reshoot any object from a controlled camera position: feed a photo, ask for a different azimuth/elevation, get the same subject from that angle. First camera-control LoRA for Qwen-Image-2.1 — the newest Qwen image base (unified generate + edit, native RGBA), where this format did not exist yet. - 36 poses: 12 azimuths (30° steps) × 4 elevations (0° eye-level / 30° elevated / 60° high-angle / 90° top-down), plus a close-up variant of each → 72 addressable framings In-distribution examples are above — but does it work on everyday photos it never saw? Dog, cat and apple from Wikimedia Commons, step-1,000 checkpoint, same prompt grammar: Golden retriever · front-right quarter view, eye-level…
Built with Qwen. Non-commercial research and evaluation use only (Qwen Research License, see LICENSE and NOTICE). An NVFP4 quantization of Qwen/Qwen-Image-2.1, made from the official BF16 weights with SVDQuant + GPTQ and calibrated on both text-to-image prompts and image edits. Both large components run on native Blackwell FP4 tensor cores, and the pipeline loads with a single DiffusionPipeline.frompretrained call. At 512x512 on an RTX PRO 6000 Blackwell it is about 1.7x faster than BF16 and needs about 40% less VRAM, and it matches BF16 closely on text-to-image, typography and edits, including a hard recolor edit that earlier 4-bit text encoders failed. trustremotecode=True loads two small…
Feed a frame, direct the next shot. This LoRA teaches Qwen-Image-2.1 cinematic scene continuation: it takes an input image and a director-style instruction and generates the next scene — camera moves, reveals, lighting shifts — while keeping the world, characters and style consistent. Chain outputs into storyboards. This is the Qwen-Image-2.1 port of the next-scene format pioneered by (313K downloads on the 2509 edit model) — retrained natively on the newest Qwen image base, where this capability did not exist yet. Companion piece to (camera control) — together they cover where the camera is and where the story goes next. ComfyUI: load in any Qwen-Image-2.1 workflow with a LoRA node…
Give Qwen Image 2.1 a picture of a person and a picture of an outfit, and it dresses the person in it. Plain Qwen 2.1 already does that, but the picture comes back a few pixels off, the face gets a little softer and the background gets touched. With this LoRA the swap happens on your picture's own frame: on 16 swaps it had never seen, the result sat 0.1 px from the original (3.6 px without it). No trigger word. Load it at strength 1.0 (at 0.5 about half of the drift comes back: 1.8 px in the same test), give the person as image 1 and the outfit as image 2, and ask for the swap: Outfit Swap workflow that has it set up. Videos on YouTube @ausbossdev. Qwen Image 2.1 can do an outfit swap…
Extends a picture in any direction: one side, two sides, a corner, three sides, all four, or a small picture on a big empty canvas. Pad the picture with flat gray #808080, give the padded canvas to Qwen Image 2.1 as the reference, and the LoRA fills the gray with a continuation of the scene while keeping the original picture where it is. side, optional LoRA](https://civitai.com/models/2964810/qwen-image-21-outpaint-auto-prompt-any-side-optional-lora) on Civitai. Videos on YouTube @ausbossdev. Pictures neither LoRA was trained on, rendered in ComfyUI with the INT8 Qwen every column, original pixels pasted back with Stitch Inpaint. Each row: box marks what goes wrong without the LoRA: visible…
GGUF builds of darrellbest/Qwen-Image-2.1-PE-I2I-Heretic, the refusal-ablated image-editing prompt rewriter: it reads the image you are editing and turns a short instruction into a detailed edit prompt. For llama.cpp and Ollama, including machines with no CUDA GPU and ARM machines such as the DGX Spark. systemprompt.txt is included and required — it defines the output contract, and the model is useless without it. The vision file. mmproj-F16.gguf is how this model reads the input image. GGUF keeps it in a separate file; both loaders take the pair and behave as one model. Ollama packs the two files into a single model, so nothing extra is needed at run time. converthftogguf.py from llama.cpp…
GGUF builds of darrellbest/Qwen-Image-2.1-PE-T2I-Heretic, the refusal-ablated text-to-image prompt rewriter: it turns a short request into a detailed prompt plus a recommended aspect ratio. For llama.cpp and Ollama, including machines with no CUDA GPU and ARM machines such as the DGX Spark. systemprompt.txt is included and required — it defines the output contract, and the model is useless without it. Ollama packs the file into a single model, so nothing extra is needed at run time. converthftogguf.py from llama.cpp (bd4f514), then llama-quantize. Converting this architecture needed a one-line fix: the Qwen converter reads mtpnumhiddenlayers with a default, but these checkpoints set it to…
Qwen Image 2.1 can upscale a small or worn picture by itself, but it overdoes it. Skin turns gritty, pores and freckles show up that were not there, faces look older, and the picture shifts a few pixels. I trained this LoRA to soften that harsh upscale. With it the result is a little softer and still detailed, it makes up less, and it stays true to the picture you gave it. I tested it on 37 pictures I had shrunk myself, so each result could be compared with the full-size picture. With the LoRA the result sat about 0.3 px off. Without it, more than 2 px. It is a redraw, not a recovery. The fine detail is made up to fit. Plain Qwen 2.1 draws more texture; this LoRA trades some of that for…
DHQ = Dual-Head Quant. Same heretic weights as quantized a different way: instead of llama.cpp's fixed rules, every tensor gets the precision that matters to what the Qwen-Image-2.1 DiT actually reads (measured per tensor with gradients through the DiT's text input), plus an importance matrix weighted the same way. Each file is exactly the size of the standard llama.cpp file of the same name, and better at every level. Tuned for Qwen-Image-2.1 only. Other models that use Qwen3-VL-8B (Ideogram 4, Boogu, JoyImageEdit) can load these files, but they were not tuned for them. Need Q5/Q6/Q8? Use the standard files in the GGUF repo — at those sizes the standard recipe is already as good. DHQ files…
NVFP4 (4-bit weight) build of Qwen-Image-2.1-Text-Encoder-Heretic, the abliterated text encoder of Qwen/Qwen-Image-2.1. File: qwen3vl8bnvfp4heretic.safetensors — 5.87 GB (from 16.33 GB bf16) Produced with Heretic directional ablation (oproj + downproj), 200 trials / 60 startup trials, knee point of the Pareto front. Measured on mlabonne/harmfulbehaviors (refusals) and mlabonne/harmlessalpaca (KL). Independently re-checked on the bf16 source: 0/20 refusals, 4/4 benign questions answered correctly. Full methodology, Pareto table and reproduction command are in the Blackwell GPUs (RTX 50xx, GB10). NVFP4 maps onto native FP4 tensor cores, so you get both the memory saving and a throughput win.…
Asymmetric W4A8 INT8 build of Qwen-Image-2.1-Text-Encoder-Heretic, the abliterated text encoder of Qwen/Qwen-Image-2.1. File: qwen3vl8bw4a8heretic.safetensors — 5.88 GB (from 16.33 GB bf16) Comfy-Org ships this format too, so it loads with the stock CLIPLoader. This is the smallest safetensors build here (6 GB). If you just want the exact format the official ComfyUI templates load by default — the "swap the filename and it works" option — use the instead; this W4A8 is the pick when you want a smaller file. Produced with Heretic directional ablation (oproj + downproj), 200 trials / 60 startup trials, knee point of the Pareto front. Measured on mlabonne/harmfulbehaviors (refusals) and…
INT8-convrot build of Qwen-Image-2.1-Text-Encoder-Heretic, the abliterated text encoder of Qwen/Qwen-Image-2.1. File: qwen3vl8bint8convrotheretic.safetensors — 8.71 GB (from 16.33 GB bf16) This is the format Comfy-Org's official Qwen-Image-2.1 templates use by default (qwen3vl8bint8convrot.safetensors). If you followed any official / community Qwen-Image-2.1 tutorial, this is a drop-in replacement — just swap the filename. It keeps every projection at INT8 (heavier than the 4-bit W4A8 build, but the most faithful to the stock template). Produced with Heretic directional ablation (oproj + downproj), 200 trials / 60 startup trials, knee point of the Pareto front. Measured on…
Qwen-Image-2.1-Turbo is an accelerated checkpoint of Qwen-Image-2.1 for text-to-image generation and image editing with 8 denoising steps. It uses the same 7B visual generation architecture and loads directly with QwenImage21Pipeline in Diffusers. The checkpoint includes its recommended sampling schedule, so it is ready to use without manually configuring the scheduler. Generation uses CFG=1 by default, and prefix KV caching reuses the text and reference-image context across denoising steps. For more details, see the GitHub repo and Qwen-Image-2.1 blog. Install a CUDA-compatible PyTorch build, then install the latest Diffusers source and the following dependencies: This checkpoint requires…
This repository provides native MLX BF16 weights for Qwen/Qwen-Image-2.1-Turbo, combined with the uncensored Heretic vision-language text encoder and the original full-precision FP32 VAE. It is ready to run directly on Apple Silicon via mflux. - 8-Step Distilled Inference: Uses the official Qwen-Image-2.1-Turbo DiT checkpoint for fast, high-quality 8-step generation without needing any external acceleration LoRA. - Official qwenturbo Sampling Sigmas: Tuned to the exact official samplesigmas grid ([1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568, 0.0]).
A community OpenVINO conversion of the official eight-step Qwen Image 2.1 Turbo checkpoint, prepared for local inference in Fern. This is not an official Qwen release or an endorsement by Qwen or Hugging Face. Install Qwen Image 2.1 Turbo from Fern → Preferences → Image models. Fern downloads the pinned OpenVINO export and matching runtime dependencies. The first generation compiles a device-specific OpenVINO cache locally; those cache files are not part of this repository. The model files occupy about 23.2 GB. Allow extra free space for temporary download data and the compiled cache. Intel GPU is the default target; CPU is an explicit fallback. The upstream Qwen Research License applies to…
GGUF quantizations of Qwen/Qwen-Image-2.1 for local image generation using the original upstream base weights. Q4KM is recommended for the best balance of size and quality. Companion model files packaged for ComfyUI: Use the model with ComfyUI and ComfyUI-GGUF. All required companion files (GGUF transformer, text encoder, and VAE) are hosted directly in this repository. Download the files and place them in their respective ComfyUI directories: 1. Install ComfyUI-GGUF: Use the maintained fork with native Qwen-Image 2.1 support by cloning leejet/ComfyUI-GGUF into your custom nodes: (Note: If you have the older city96/ComfyUI-GGUF installed and encounter an Unknown model architecture! error…
Model Collections
Hand-picked starting points, each with the reason it exists.
Collection · 4 entries
Embedding models for retrieval
Sentence and document embedding models used to build retrieval systems. Dimension and sequence length matter more than size here, and both come from the publisher.
Collection · 4 entries
Models that fit on one accelerator
Models whose publisher-reported parameter count puts them within reach of a single accelerator at common precisions. Memory needed depends on precision and serving configuration, so treat the parameter count as the starting point, not the answer.
Collection · 6 entries
Open-weight text models worth knowing
Widely used open-weight language models, chosen because each one is a distinct family rather than a variant of the one above it. Selection, not a ranking.
Collection · 3 entries
Speech and audio models
Recognition and synthesis models, grouped so the two directions are easy to compare.
Open-Weight Models Explained
What is an open-weight model?
An AI model whose trained weights are published for anyone to download, so it can be run, tested and fine-tuned on hardware the user controls.
Is an open-weight model the same as open source?
Not always. Open weights means the trained model can be downloaded. Open source usually also means the training code and data are available and the license allows broad reuse. Many open-weight models release the weights only.
Can I use an open-weight model commercially?
It depends on the license. Apache 2.0 and MIT allow commercial use. Other licenses limit it, for example to non-commercial use or below a set number of users. Every model page here shows its license.
How much memory does an open-weight model need?
About two bytes per parameter at 16-bit precision, so a 7-billion-parameter model needs roughly 14 GB for its weights, plus memory for the context it processes. Each model page lists its parameter count and the size of its files.
Related SAVRN Research
The hub sits beside SAVRN's market data and infrastructure research: what models cost to run, and what it takes to run them.
SAVRN Index
What open models cost to run
The same open-weight model priced by every host that serves it, per million tokens.
Research Hub
Data center trackers and maps
Moratoriums, permits, power, water and capital behind the facilities that run these models.
Method
How the Model Hub is built
Sources, evidence labels, refresh behaviour, and the limits of every comparison here.