https://vramready.com/new/2026-09-25T00:00:00ZVRAM Readyhttps://vramready.com/models/qwen-image-2-1/2026-09-01T00:00:00Z

September 2026: a new 7B model that does text-to-image and editing in one, with transparent-background (RGBA) output. Far lighter than the 20B Qwen-Image. Fits from 6 GB of VRAM; 8-bit or better from 10 GB.

https://vramready.com/models/wan-animate-2/2026-07-01T00:00:00Z

The July 2026 successor to Wan 2.2 Animate, adding text-driven camera and viewpoint control. The distilled variant runs in about 10 steps. Fits from 12 GB of VRAM; 8-bit or better from 22 GB.

https://vramready.com/models/minimax-h3-pruned/2026-07-01T00:00:00Z

An official, smaller cut of MiniMax H3 published by Comfy-Org, made for cards that cannot hold the full 33B model. Fits from 15 GB of VRAM; 8-bit or better from 27 GB.

https://vramready.com/models/minimax-h3/2026-07-01T00:00:00Z

The biggest open video release of 2026 and the most downloaded ComfyUI repo of the year: 33B parameters, 4–15 second clips with native stereo sound. Two variants (first/last-frame and reference-to-video); load one. Fits from 22 GB of VRAM; 8-bit or better from 40 GB.

https://vramready.com/models/mage-flow/2026-07-01T00:00:00Z

Microsoft's 4B native-resolution model from July 2026, with separate Turbo (4-step) and Edit checkpoints of the same size. Light enough for most GPUs. Fits from 6 GB of VRAM; 8-bit or better from 6 GB.

https://vramready.com/models/ltx-2-5/2026-07-01T00:00:00Z

The July 2026 LTX, with a custom Gemma 4 12B text encoder. GGUF sizes differ a lot between uploaders because they mix precisions differently; figures here use realrebelai's distilled files. Fits from 14 GB of VRAM; 8-bit or better from 29 GB.

https://vramready.com/models/scail-2/2026-06-01T00:00:00Z

June 2026 character animation and replacement built on Wan 2.1 14B: a reference image driven by the motion of another video, without pose skeletons. Native ComfyUI nodes. Fits from 13 GB of VRAM; 8-bit or better from 22 GB.

https://vramready.com/models/krea-2/2026-06-01T00:00:00Z

Krea's own 12B model from June 2026, one of the most downloaded ComfyUI releases of the year. Turbo is the 8-step distilled version; Raw is the base for fine-tuning, same size. Fits from 8 GB of VRAM; 8-bit or better from 16 GB.

https://vramready.com/models/ideogram-4/2026-06-01T00:00:00Z

Ideogram's open-weights model (June 2026, non-commercial). It uses two transformers at once — a main one and an "unconditional" one — so it needs about twice the memory of one file. Fits from 15 GB of VRAM; 8-bit or better from 21 GB.

https://vramready.com/models/boogu-image/2026-06-01T00:00:00Z

A 10B model from June 2026 that generates and edits in one. Turbo needs about 4 steps; Base and the separate Edit checkpoints are the same size. Fits from 11 GB of VRAM; 8-bit or better from 13 GB.

https://vramready.com/models/hidream-o1-image/2026-05-01T00:00:00Z

A 2026 all-in-one 8.8B model built on Qwen3-VL: no separate text encoder and no VAE. It generates, edits and personalises up to 2048×2048. Fits from 11 GB of VRAM; 8-bit or better from 11 GB.

https://vramready.com/models/ernie-image/2026-04-01T00:00:00Z

Baidu's 8B model from April 2026, strong at rendering text in images. Turbo runs in about 8 steps; Base takes 50. Fits from 6 GB of VRAM; 8-bit or better from 11 GB.

https://vramready.com/models/ltx-2-3/2026-03-01T00:00:00Z

The March 2026 update of LTX-2, grown to 22B parameters. Fits from 14 GB of VRAM; 8-bit or better from 28 GB.

https://vramready.com/models/z-image/2026-01-01T00:00:00Z

The undistilled base of Z-Image Turbo: same 6B size, but 28–50 steps with CFG. Better for LoRA training and negative prompts. Fits from 7 GB of VRAM; 8-bit or better from 9 GB.

https://vramready.com/models/ltx-2/2026-01-01T00:00:00Z

Lightricks' open audio+video model. Generates sound together with the picture. Its text encoder is Gemma 3 12B. Fits from 13 GB of VRAM; 8-bit or better from 26 GB.

https://vramready.com/models/flux-2-klein-9b/2026-01-01T00:00:00Z

A distilled 9B FLUX.2 for fast generation and editing in a few steps. Much lighter than FLUX.2 dev. Fits from 7 GB of VRAM; 8-bit or better from 12 GB.

https://vramready.com/models/flux-2-klein-4b/2026-01-01T00:00:00Z

The smallest FLUX.2: 4B parameters, Apache-2.0 licence, a few steps per image. Fits almost any modern card. Fits from 4 GB of VRAM; 8-bit or better from 6 GB.

https://vramready.com/models/qwen-image-edit/2025-12-01T00:00:00Z

The editing version of Qwen-Image, release 2511. Same 20B size and the same text encoder as Qwen-Image. Fits from 11 GB of VRAM; 8-bit or better from 24 GB.

https://vramready.com/models/z-image-turbo/2025-11-01T00:00:00Z

A 6B, 8-step model from Alibaba's Tongyi lab and one of the most downloaded ComfyUI models of the past year. Fast and light. Fits from 6 GB of VRAM; 8-bit or better from 9 GB.

https://vramready.com/models/hunyuanvideo-1-5/2025-11-01T00:00:00Z

Tencent's 8.3B video model from late 2025, much lighter than the original 13B HunyuanVideo. The FP8 file is the cfg-distilled variant. Fits from 10 GB of VRAM; 8-bit or better from 13 GB.

https://vramready.com/models/flux-2-dev/2025-11-01T00:00:00Z

The 32B second-generation FLUX. Very capable and very heavy: even 4-bit files are around 20 GB, and the text encoder is a 24B language model. Fits from 17 GB of VRAM; 8-bit or better from 39 GB.

https://vramready.com/models/wan-2-2-animate-14b/2025-09-01T00:00:00Z

Character animation and replacement: a reference image of a person is driven by the motion and face of another video. Fits from 12 GB of VRAM; 8-bit or better from 23 GB.

https://vramready.com/models/hunyuanimage-2-1/2025-09-01T00:00:00Z

Tencent's 17B model that generates natively at 2K. Its GGUF files come from calcuis and need calcuis's own GGUF loader node. Fits from 11 GB of VRAM; 8-bit or better from 21 GB.

https://vramready.com/models/wan-2-2-s2v-14b/2025-08-01T00:00:00Z

Sound-to-video: a talking or singing person from one image plus an audio track. Fits from 15 GB of VRAM; 8-bit or better from 22 GB.

https://vramready.com/models/qwen-image/2025-08-01T00:00:00Z

Alibaba's 20B text-to-image model, known for rendering long text and posters well. Big: the 8-bit file alone is over 20 GB. Fits from 10 GB of VRAM; 8-bit or better from 24 GB.

https://vramready.com/models/chroma1-hd/2025-08-01T00:00:00Z

An 8.9B community model built from FLUX.1 schnell, Apache-2.0 licensed, with a big fine-tuning scene. Fits from 6 GB of VRAM; 8-bit or better from 12 GB.

https://vramready.com/models/wan-2-2-5b/2025-07-01T00:00:00Z

A single dense 5B model for text- or image-to-video at 720p. The friendliest Wan for 8–12 GB cards. Fits from 6 GB of VRAM; 8-bit or better from 10 GB.

https://vramready.com/models/wan-2-2-t2v-14b/2025-07-01T00:00:00Z

Wan 2.2 splits the work between two 14B models: one for the early, noisy steps and one for the late ones. Only one needs to be in VRAM at a time, but both must fit in system RAM. Fits from 10 GB of VRAM; 8-bit or better from 19 GB.

https://vramready.com/models/wan-2-2-i2v-14b/2025-07-01T00:00:00Z

The image-to-video Wan 2.2, and the most downloaded Wan 2.2 GGUF. Same two-model setup as the text-to-video version. Fits from 10 GB of VRAM; 8-bit or better from 19 GB.

https://vramready.com/models/ltx-video-13b/2025-07-01T00:00:00Z

The last of the pre-LTX-2 video models from Lightricks (July 2025). No audio, T5 text encoder, and much lighter than LTX-2. Fits from 10 GB of VRAM; 8-bit or better from 19 GB.