2026-09-26 · America/Los_Angeles · 社区动态 · #11

I know there are already 5 of these, but I made a LLM Pareto frontier graph

The main motivation was my coworkers who hype every rumoured model drop and token limit reset like it's the second coming. Now I at least have a chart to point at and a proper list of all rumored upcoming releases. Since all existing frontier listings looked bad and didn't have latest models I just redid the concept. Stuff that surprised me: - MiniMax H3 scores way higher in video gen than I expected - Outside of that, open weight models get pretty consistently destroyed in image and video gen. The gap is way bigger there than in text Happy to get any comments really. Trying to get a honest…

热度 46.0 / 100；排名与评分保留该期记录。