Run models

Find an uncensored AI model in three questions

By the SpicyAPI team · Updated · Test data as of · How we test

The picker filters 130 tested models down to the ones that fit your project, then sorts them by what you care about. It only recommends models that delivered at least 70% of our NSFW test runs.

1. What are you making?
2. What do you start from?
3. What matters most?

Your matches

  1. Best match

    Wan 2.6 Flash

    19 of 19 test runs delivered · blind test 69.2 · from $0.022/sec at 720p

    Run Wan 2.6 Flash on SpicyAPI Why it ranks here →

  2. Alternative

    MiniMax H3 Singularity LoRA

    38 of 38 test runs delivered · blind test 58.3 · from $0.06/sec at 480p

    Run MiniMax H3 Singularity LoRA on SpicyAPI

  3. Alternative

    Wan 2.7 Spicy

    19 of 19 test runs delivered · blind test 61 · from $0.12/sec at 720p

    Run Wan 2.7 Spicy on SpicyAPI Why it ranks here →

How the picker decides

The first two answers narrow the pool: a video project with a still image only considers models that accept a start frame, and choosing “my own LoRA file” limits the pool to builds that load LoRA weights by link. The third answer sets the order. Range sorts by Freedom Score, which also penalises levels that were never tested; quality sorts by blind-test score; price sorts by the lowest listed price, which is per second for video and per image for stills.

Your answers stay in your browser. Nothing is sent to us or to SpicyAPI until you click through to a model.

Questions

Why is my favourite model missing?

Either it delivered less than 70% of the test runs, or it does not accept the input you chose. The filter tracker lists every model with its full results.

Is the cheapest option good enough?

Often for drafts. Check the blind-test score in the result: a low score usually means more re-runs, which can cost more than starting on a better model.