Sana
Efficient diffusion image generation that fits on one GPU.
About
Sana is NVIDIA's text-to-image diffusion model built around a linear diffusion transformer, designed to generate high-resolution images fast on a single GPU instead of a cluster. The 1.6B model became a popular pick for local and ComfyUI workflows, and the family later grew to include a sprint-distilled variant and a video model. Weights ship under NVIDIA's Open Model License, which allows commercial use but is revocable and not OSI-open, and the bundled Gemma text encoder adds its own terms. Prompt understanding is decent rather than class-leading, and the CUDA-focused design runs slower on non-NVIDIA hardware.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Authors position Sana-1.6B as competitive with much larger diffusion models at a fraction of the latency (NVIDIA).
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Scored 7/10 (verified tier, metrics-v1). Evidence: NVIDIA NVLabs; 1.6B model popular for local/ComfyUI image generation; NVIDIA Open Model License: commercial use allowed but revocable, not OSI-open; Bundled Gemma-2 text encoder adds Gemma terms on top; CUDA-focused; slower off NVIDIA hardware; prompt understanding decent not elite. License: NVIDIA Open Model License (commercial allowed, revocable); confirm on the official card..
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 7.0/10
- Maintenance · recency, releases, issue load
- 7.0/10
- Openness · license + docs
- 6.0/10
- Manual review · sanity read + red-flag check
- 6.0/10
- · NVIDIA NVLabs; 1.6B model popular for local/ComfyUI image generation
- · NVIDIA Open Model License: commercial use allowed but revocable, not OSI-open
- · Bundled Gemma-2 text encoder adds Gemma terms on top
- · CUDA-focused; slower off NVIDIA hardware; prompt understanding decent not elite