ComfyUI INT4 Quantized Model Collection
Winougan · reddit · 2026-07-12
The author uploaded a batch of INT4 quantized ComfyUI models on Hugging Face, along with workflows and examples, emphasizing "free to use".
Key Info
- Aim to run models on lower VRAM; tested on RTX 3070 Ti and RTX 4090.
- Recommended environment: ComfyUI nightly, PyTorch 2.12, Python 3.13, cu132, Triton 3.8, FlashAttention 2, SageAttention 2.
- Performance claims:
- INT8 offers 25% improvement over BF16;
- INT4 offers 40%–50% improvement.
- Quality: INT8 nearly identical to BF16, INT4 close to FP8.
Uploaded/To be uploaded
- Uploaded: SeedVR 7b INT4, Gemma 3 12b INT4, Sulphur 2 Base INT4.
- Planned: Krea 2 Turbo, Klein9b, Z-Image Turbo, some Illustrious XL models, plus custom fine-tuned versions.
Related event: Multiple Image Models Get INT4 Quantization to Lower VRAM Requirements(2 posts)→
More from Infra
- SpaceX strikes $40B deal with Apollo Global to lease AI chips and data center hardware — Polymarket · 2026-10-26
- India Needs at Least 5GW of Compute by 2028 to Stay Relevant, User Argues — EAccelerate_42 · 2026-10-26
- SpaceX AI revenue up 830%, Claude designs 318K antivenom proteins in 64 hours — PeterDiamandis · 2026-10-26
- Meta plans up to $95 billion stock sale to fund AI buildout — The Decoder · 2026-10-26
- Pandora M1: first prototype of a physical device for distributed GPU monitoring — tarantulae · 2026-10-26
- INT21's AI agent swarm rewrote Drama 3's inference runtime in Rust within one hour, beating SGLang-Omni — bingxu_ · 2026-10-26