ImageBench V1 —

192 evaluations across 6 categories

Benchmark V1 verdicts are produced by VLM judges and can contain mistakes. Treat PASS/FAIL labels as machine-assisted assessments, and inspect the images yourself. Learn more about the methodology.

Generation Details

Source-backed model context, size, cost, and request settings for this ImageBench V1 run.

local/tld-101m

Local

Transformer Latent Diffusion (TLD) is a minimal 101M-parameter latent diffusion transformer by apapiu (MIT), trained for about 32 hours on one A100. It conditions on a pooled CLIP ViT-L/14 embedding, samples with DPM-Solver++(2M) and outputs 256x256 images; it is a small educational/research model.

Maker: apapiu (independent)
Family: Transformer Latent Diffusion
Model Size: 101M
high
Cost: local run; no API price
not_applicable
Run Target: gx10/tld-101m
Effective Request: response_format: b64_json · size: 256x256 · num_inference_steps: 15 · guidance_scale: 6 · seed: 7
Sources: apapiu/transformer_latent_diffusion (GitHub)apapiu/small_ldt checkpoint