MiniMax M3.1 Flash beats DeepSeek V4.1 Flash, Opus 5.5 and GLM 5.3 Flash on a cloth-wrinkling prompt
On a paper-and-fabric bending test, M3.1 Flash kept structure, texture and motion best; DeepSeek showed sharp folds, Opus's paper was too soft, GLM too rigid, and M3.1 Flash had the best token value.
tested MiniMax M3.1 Flash against DeepSeek V4.1 Flash, Claude Opus 5.5 and GLM 5.3 Flash on the same prompt the task: simulate paper and elastic fabric bending, wrinkling and deforming under pressure M3.1 Flash delivered the strongest result: - structural integrity: the receipt holds its shape through bending and wrinkling, and the printed text stays clear and legible - material detail: the fabric keeps fine texture and deforms smoothly under pressure - motion control: no sharp or unnatural folds once things start moving the others fell short in motion: - DeepSeek V4.1 Flash looked fine static, but large sharp folds appeared once the materials moved - Claude Opus 5.5 had strong lighting and composition, but the paper twisted like it was too thin and soft - GLM 5.3 Flash kept the shapes, but the paper stayed too rigid and the fabric texture repeated what stands out is the efficiency. M3.1 Flash delivers this level of detail and consistency at a low token cost, and its tokens were the best value of the four in this test stronger results for every token spent is what makes a model practical for real production work, not just impressive in a demo what do you think?