Best Image Captioning Models

Image captioning is the task of generating a natural language description of an image. Architecturally it is an encoder-decoder problem: a vision encoder turns the image into features, and a language decoder writes text conditioned on them, which is exactly the structure of modern vision language models. Variants include short captions (one summary sentence), detailed captions, and dense captioning that describes individual regions. Automatic metrics like CIDEr and BLEU exist but mostly reward phrasing overlap with reference captions, so human or model-based judgment is common in practice. Captioning is used for alt text and accessibility, media library indexing, and making image collections searchable by content. This page lists 68 image captioning models, including 29 open-weight options you can self-host; 62 of them run live in the Playground so you can test them on your own images.

It depends on your task and constraints. For fixed categories in production, a model fine-tuned on your own data typically beats any general-purpose model. Compare the image captioning models on this page and try them on your own image to see which fits.

Yes. 29 of the 68 image captioning models here are open-weight (for example Kimi K3, Gemma 4 12B, and Qwen3.6 27B), free to self-host under their licenses (Modified MIT, Apache 2.0, and MIT).

Yes. You can run 62 of them in the Roboflow Playground for free. Upload an image and compare the models' output side by side, no setup required.