zsxkib/blip-3

Blip 3 / XGen-MM, Answers questions about images ({blip3,xgen-mm}-phi3-mini-base-r-v1)

Capabilities

Reference ImagesSystem PromptMax TokensTop-P

Cost

Community model (estimated from hardware time)

Input Parameters

Name	Type	Description	Default	Constraints
`image`*	string(uri)	Input image	`—`	—
`caption`	boolean	Select if you want to generate image captions instead of asking questions	`false`	—
`context`	string	Optional - previous questions and answers to be used as context for answering current question	`—`	—
`do_sample`	boolean	Whether to use sampling or not	`false`	—
`length_penalty`	number	The parameter for length penalty	`1`	min: 0
`max_new_tokens`	integer	Maximum number of new tokens to generate	`768`	min: 1, max: 2048
`num_beams`	integer	Number of beams for beam search	`1`	min: 1, max: 10
`question`	string	Question to ask about this image	`"What is shown in the image?"`	—
`repetition_penalty`	number	The parameter for repetition penalty	`1`	min: 0
`system_prompt`	string	System prompt	`"A chat between a curious user and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the user's questions."`	—
`temperature`	number	Temperature for use with nucleus sampling	`1`	min: 0.5, max: 1
`top_k`	integer	The number of highest probability vocabulary tokens to keep for top-k sampling	`50`	min: 1
`top_p`	number	The cumulative probability threshold for top-p sampling	`1`	min: 0, max: 1

imagerequiredstring

Input image

captionboolean

Select if you want to generate image captions instead of asking questions

Default: false

contextstring

Optional - previous questions and answers to be used as context for answering current question

do_sampleboolean

Whether to use sampling or not

Default: false

length_penaltynumber

The parameter for length penalty

Default: 1min: 0

max_new_tokensinteger

Maximum number of new tokens to generate

Default: 768min: 1, max: 2048

num_beamsinteger

Number of beams for beam search

Default: 1min: 1, max: 10

questionstring

Question to ask about this image

Default: "What is shown in the image?"

repetition_penaltynumber

The parameter for repetition penalty

Default: 1min: 0

system_promptstring

System prompt

Default:

"A chat between a curious user and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the user's questions."

temperaturenumber

Temperature for use with nucleus sampling

Default: 1min: 0.5, max: 1

top_kinteger

The number of highest probability vocabulary tokens to keep for top-k sampling

Default: 50min: 1

top_pnumber

The cumulative probability threshold for top-p sampling

Default: 1min: 0, max: 1

Version: 499bec581d8fUpdated: 7/25/20261.3M runs