|
gptkbp:instanceOf
|
large language model
|
|
gptkbp:abilities
|
translator
code generation
question answering
summarization
text generation
image input (multimodal)
|
|
gptkbp:architecture
|
gptkb:transformation
|
|
gptkbp:availableOn
|
subscription
API access
|
|
gptkbp:citation
|
https://arxiv.org/abs/2303.08774
|
|
gptkbp:context
|
32k tokens (variant)
8k tokens
|
|
gptkbp:developedBy
|
gptkb:OpenAI
|
|
gptkbp:features
|
improved over GPT-3.5
|
|
https://www.w3.org/2000/01/rdf-schema#label
|
GPT-4
|
|
gptkbp:input
|
gptkb:text
images (limited)
|
|
gptkbp:license
|
proprietary
|
|
gptkbp:mobileAppAvailable
|
gptkb:OpenAI_API
|
|
gptkbp:multimodal
|
yes
|
|
gptkbp:notableFor
|
better factuality
better reasoning
fewer hallucinations
higher accuracy than GPT-3.5
improved steerability
multimodal capabilities
reduced bias
|
|
gptkbp:notableUser
|
gptkb:Stripe
gptkb:Morgan_Stanley
gptkb:Duolingo
gptkb:Khan_Academy
gptkb:Be_My_Eyes
|
|
gptkbp:output
|
gptkb:text
|
|
gptkbp:parameter
|
undisclosed
|
|
gptkbp:passesThrough
|
gptkb:LSAT
gptkb:SAT
gptkb:AP_exams
bar exam (top 10%)
|
|
gptkbp:predecessor
|
gptkb:GPT-3.5
|
|
gptkbp:releaseDate
|
2023-03-14
|
|
gptkbp:successor
|
gptkb:GPT-4o
gptkb:GPT-5_(rumored)
|
|
gptkbp:supportsLanguage
|
multiple languages
|
|
gptkbp:trainer
|
internet text
|
|
gptkbp:trainingDataCutoff
|
2023-04
|
|
gptkbp:usedIn
|
gptkb:Bing_Chat
gptkb:ChatGPT_Plus
gptkb:Microsoft_Copilot
|
|
gptkbp:bfsParent
|
gptkb:Upanishads
|
|
gptkbp:bfsLayer
|
4
|