daanelson/minigpt-4

A model which generates text in response to an input image and prompt.

Capabilities

Reference Images Max Tokens Top-P

Community model (estimated from hardware time)

Name	Type	Description	Default	Constraints
`image` *	string (uri)	Image to discuss	`—`	—
`prompt` *	string	Prompt for mini-gpt4 regarding input image	`—`	—
`max_length`	integer	Total length of prompt and output in tokens	`4000`	min: 1
`max_new_tokens`	integer	Maximum number of new tokens to generate	`3000`	min: 1
`num_beams`	integer	Number of beams for beam search decoding	`3`	min: 1, max: 10
`repetition_penalty`	number	Penalty for repeated words in generated text; 1 is no penalty, values greater than 1 discourage repetition, less than 1 encourage it.	`1`	min: 0.01, max: 5
`temperature`	number	Temperature for generating tokens, lower = more predictable results	`1`	min: 0.01, max: 2
`top_p`	number	Sample from the top p percent most likely tokens	`0.9`	min: 0, max: 1

image required string

Image to discuss

prompt required string

Prompt for mini-gpt4 regarding input image

max_length integer

Total length of prompt and output in tokens

Default: 4000 min: 1

max_new_tokens integer

Maximum number of new tokens to generate

Default: 3000 min: 1

num_beams integer

Number of beams for beam search decoding

Default: 3 min: 1, max: 10

repetition_penalty number

Penalty for repeated words in generated text; 1 is no penalty, values greater than 1 discourage repetition, less than 1 encourage it.

Default: 1 min: 0.01, max: 5

temperature number

Temperature for generating tokens, lower = more predictable results

Default: 1 min: 0.01, max: 2

top_p number

Sample from the top p percent most likely tokens

Default: 0.9 min: 0, max: 1

Version: e447a8583cff Updated: 2/26/2026 1.8M runs