{"passport":{"unfragile":{"@version":"1.0","version":"2026-05","artifact":{"id":"openrouter-openai-gpt-4","slug":"openai-gpt-4","name":"OpenAI: GPT-4","type":"model","url":"https://openrouter.ai/models/openai~gpt-4","page_url":"https://unfragile.ai/openai-gpt-4","categories":["chatbots-assistants"],"tags":["openai","api-access","text"],"pricing":{"model":"paid","free":false,"starting_price":"$3.00e-5 per prompt token"},"status":"active","verified":false},"capabilities":[{"id":"openrouter-openai-gpt-4__cap_0","uri":"capability://image.visual.multimodal.reasoning.with.vision.and.text.integration","name":"multimodal reasoning with vision and text integration","description":"GPT-4 processes both text and image inputs through a unified transformer architecture, using vision encoders to embed images into the same token space as text, enabling joint reasoning across modalities. The model performs end-to-end training on interleaved image-text sequences, allowing it to answer questions about images, extract text from screenshots, analyze diagrams, and reason about visual content without separate vision-language alignment layers.","intents":["I need to analyze a screenshot and extract structured data from it","I want to ask questions about an image and get detailed explanations","I need to describe what's happening in a diagram or chart and get insights","I want to extract text from a PDF or image and process it further"],"best_for":["developers building document processing pipelines","teams automating visual QA and screenshot analysis","builders creating accessibility tools that describe images"],"limitations":["Image resolution capped at ~2000x2000 pixels; larger images are downsampled, losing fine detail","Cannot process video or animated content — only static images","Vision performance degrades on highly stylized or artistic images vs photorealistic content","No real-time video stream processing; requires discrete image submissions"],"requires":["OpenAI API key with GPT-4 Vision access enabled","Image input as base64-encoded data or URL (JPEG, PNG, GIF, WebP supported)","HTTP client capable of multipart form submission"],"input_types":["text","image (JPEG, PNG, GIF, WebP)","image URLs (publicly accessible)"],"output_types":["text","structured JSON (via prompt engineering)","code snippets"],"categories":["image-visual","text-generation-language"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"openrouter-openai-gpt-4__cap_1","uri":"capability://planning.reasoning.chain.of.thought.reasoning.with.step.by.step.decomposition","name":"chain-of-thought reasoning with step-by-step decomposition","description":"GPT-4 implements implicit chain-of-thought reasoning through its training on reasoning-heavy datasets, allowing it to generate intermediate reasoning steps before producing final answers. When prompted to 'think step by step', the model allocates more compute tokens to exploring solution paths, backtracking when needed, and validating intermediate conclusions before committing to outputs. This is achieved through instruction-tuning on datasets where reasoning traces precede answers.","intents":["I need to solve a complex math problem and see the working","I want the model to explain its reasoning for a decision or classification","I need to debug why a piece of code isn't working by walking through logic","I want to verify a model's answer by inspecting its reasoning process"],"best_for":["educators building tutoring systems that need to show work","developers debugging LLM behavior in production","teams building verification systems that need interpretability"],"limitations":["Reasoning quality is prompt-dependent; 'think step by step' is a heuristic, not guaranteed reasoning","No access to intermediate reasoning tokens — only final text output is visible","Reasoning traces can be verbose, increasing token consumption by 2-5x vs direct answers","No formal proof of correctness; reasoning can contain plausible-sounding but incorrect steps"],"requires":["OpenAI API key with GPT-4 access","Prompt engineering to trigger reasoning (e.g., 'Let's think step by step')","Sufficient token budget for longer outputs (typically 500-2000 tokens for reasoning traces)"],"input_types":["text","structured prompts with reasoning instructions"],"output_types":["text with reasoning traces","step-by-step explanations","intermediate conclusions"],"categories":["planning-reasoning","text-generation-language"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"openrouter-openai-gpt-4__cap_10","uri":"capability://data.processing.analysis.sentiment.analysis.and.text.classification.with.custom.categories","name":"sentiment analysis and text classification with custom categories","description":"GPT-4 classifies text into sentiment categories (positive, negative, neutral) or custom categories by learning classification patterns through instruction-tuning on labeled examples. The model uses transformer attention to identify sentiment-bearing words, context, and implicit meaning, enabling nuanced classification that handles sarcasm, mixed sentiment, and domain-specific language. Classification can be zero-shot (no examples) or few-shot (with examples), with few-shot improving accuracy.","intents":["I need to classify customer reviews or feedback by sentiment","I want to categorize text into custom categories (e.g., product types, issue types)","I need to detect intent in user messages (e.g., complaint, question, praise)","I want to analyze social media posts or comments for sentiment and topics"],"best_for":["developers building content moderation or feedback analysis systems","teams analyzing customer sentiment at scale","builders creating intent detection for chatbots or customer support"],"limitations":["Classification is probabilistic; no confidence scores or uncertainty estimates returned","Performance on custom categories depends on clarity of category definitions and number of examples","Sarcasm and implicit sentiment are sometimes misclassified","No built-in handling of multilingual text; language must be specified for best results","Batch classification requires multiple API calls; no native batch processing endpoint"],"requires":["OpenAI API key with GPT-4 access","Text to classify","Category definitions (for custom classification)","Optional: few-shot examples for improved accuracy"],"input_types":["text (review, comment, message)","category definitions or examples"],"output_types":["text (category label)","structured JSON (with category and confidence)"],"categories":["data-processing-analysis","text-generation-language"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"openrouter-openai-gpt-4__cap_11","uri":"capability://data.processing.analysis.structured.data.extraction.from.unstructured.text","name":"structured data extraction from unstructured text","description":"GPT-4 extracts structured information (entities, relationships, attributes) from unstructured text by learning extraction patterns through instruction-tuning on examples where text is paired with structured outputs (JSON, tables). The model uses transformer attention to identify relevant spans of text, map them to schema fields, and format outputs according to specified schemas. Extraction can be guided by providing a target schema or examples of desired output format.","intents":["I need to extract entities (names, dates, locations) from documents","I want to parse unstructured data into a structured database format","I need to extract key-value pairs or attributes from text","I want to convert natural language descriptions into structured data (e.g., product specs)"],"best_for":["developers building data pipelines that ingest unstructured text","teams automating document processing and data entry","builders creating knowledge extraction systems"],"limitations":["Extraction accuracy depends on schema clarity and text quality; ambiguous schemas produce inconsistent results","No built-in validation that extracted data conforms to schema; requires post-processing validation","Hallucination risk: model may invent data not present in the source text","Performance degrades on domain-specific terminology or formats not well-represented in training data","No built-in handling of nested or complex schemas; deeply nested structures may be flattened or lost"],"requires":["OpenAI API key with GPT-4 access","Unstructured text input","Target schema (JSON schema, table format, or examples)","Optional: few-shot examples for improved accuracy"],"input_types":["text (unstructured document, article, description)","schema definition (JSON schema or examples)"],"output_types":["JSON (structured data conforming to schema)","CSV or table format","key-value pairs"],"categories":["data-processing-analysis","text-generation-language"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"openrouter-openai-gpt-4__cap_12","uri":"capability://planning.reasoning.prompt.optimization.and.few.shot.learning.with.in.context.examples","name":"prompt optimization and few-shot learning with in-context examples","description":"GPT-4 improves task performance through few-shot learning by conditioning on examples of input-output pairs provided in the prompt. The model uses transformer attention to recognize patterns in the examples and apply them to new inputs, enabling task adaptation without fine-tuning. Few-shot learning is particularly effective for custom tasks, domain-specific language, and non-standard output formats. Performance typically improves with 2-5 examples; diminishing returns occur beyond 10 examples.","intents":["I want to adapt GPT-4 to a custom task without fine-tuning","I need to teach the model a specific output format through examples","I want to improve accuracy on domain-specific tasks by providing examples","I need to handle non-standard or custom language patterns"],"best_for":["developers building task-specific applications without fine-tuning infrastructure","teams rapidly prototyping new use cases with minimal data","builders creating customizable systems where users can define tasks via examples"],"limitations":["Few-shot learning quality depends on example quality and representativeness; poor examples degrade performance","Context window limits number of examples; typically 2-10 examples fit in available context","No learning persistence; examples must be provided with every request","Performance gains from few-shot learning are smaller than fine-tuning on large datasets","Example selection is manual; no automated mechanism to identify optimal examples"],"requires":["OpenAI API key with GPT-4 access","2-10 representative examples of input-output pairs","Clear task description or instruction"],"input_types":["text (task description)","examples (input-output pairs demonstrating desired behavior)"],"output_types":["text (output following the pattern demonstrated in examples)","structured data (if examples show structured format)"],"categories":["planning-reasoning","text-generation-language"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"openrouter-openai-gpt-4__cap_2","uri":"capability://code.generation.editing.code.generation.and.completion.with.context.aware.synthesis","name":"code generation and completion with context-aware synthesis","description":"GPT-4 generates code across 50+ programming languages by learning patterns from public code repositories and documentation during pretraining. It uses transformer attention to track variable scope, function signatures, and import dependencies across files, enabling it to generate syntactically correct and semantically coherent code snippets. The model can complete partial functions, generate boilerplate, refactor existing code, and explain code logic through instruction-tuning on code-explanation pairs.","intents":["I need to generate a function that solves a specific problem described in English","I want to complete a partially-written function or class","I need to refactor existing code to improve readability or performance","I want to understand what a piece of code does and how to modify it"],"best_for":["solo developers accelerating prototyping and boilerplate generation","teams using GPT-4 as a pair programmer for code review and suggestions","educators teaching programming by having students explain generated code"],"limitations":["No real-time compilation or execution; generated code may have syntax errors or logical bugs","Context window limited to ~8,000 tokens (GPT-4 standard) or ~32,000 (GPT-4 Turbo); cannot process entire large codebases","No awareness of project-specific conventions, internal libraries, or custom frameworks unless explicitly provided in context","Performance degrades on domain-specific languages (DSLs) or languages with <1% representation in training data"],"requires":["OpenAI API key with GPT-4 access","Code context provided as text (function signatures, imports, docstrings)","Language specification in prompt (e.g., 'Write this in Python 3.11')"],"input_types":["text (natural language description)","code snippets (partial functions, class definitions)","pseudocode or algorithm descriptions"],"output_types":["code (Python, JavaScript, Java, C++, Go, Rust, etc.)","code with inline comments","refactored code with explanations"],"categories":["code-generation-editing","text-generation-language"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"openrouter-openai-gpt-4__cap_3","uri":"capability://tool.use.integration.function.calling.with.schema.based.tool.binding","name":"function calling with schema-based tool binding","description":"GPT-4 supports structured function calling by accepting a JSON schema of available functions and returning structured JSON objects specifying which function to call and with what arguments. The model learns to map natural language requests to function calls through instruction-tuning on examples where user intents are paired with function invocations. This enables deterministic tool orchestration without parsing natural language outputs, as the model directly outputs structured data conforming to the provided schema.","intents":["I need to route user requests to specific backend APIs based on intent","I want to build an agent that calls multiple tools in sequence to solve a problem","I need to extract structured data (e.g., function arguments) from user input","I want to integrate GPT-4 with my existing API layer without custom parsing"],"best_for":["developers building LLM agents with deterministic tool calling","teams integrating GPT-4 into existing API-driven architectures","builders creating multi-step workflows that require tool orchestration"],"limitations":["Function calling is non-deterministic; model may hallucinate function names or arguments not in the schema","No built-in error handling or retry logic if a function call fails; requires external orchestration","Schema complexity is limited by context window; deeply nested or very large schemas may be truncated","No native support for streaming function calls; entire response must be buffered before parsing"],"requires":["OpenAI API key with GPT-4 access","Function schema defined in JSON Schema format (OpenAI's function_calling format)","HTTP client capable of parsing structured JSON responses"],"input_types":["text (user request or query)","JSON schema (function definitions with parameters)"],"output_types":["JSON (function name and arguments)","structured function call objects"],"categories":["tool-use-integration","planning-reasoning"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"openrouter-openai-gpt-4__cap_4","uri":"capability://text.generation.language.knowledge.synthesis.and.question.answering.with.broad.domain.coverage","name":"knowledge synthesis and question answering with broad domain coverage","description":"GPT-4 answers questions across diverse domains (science, history, law, medicine, programming) by leveraging knowledge learned during pretraining on internet text, books, and academic papers up to April 2023. The model uses transformer attention to retrieve relevant knowledge from its parameters and synthesize coherent answers, combining multiple facts and reasoning steps. Knowledge is implicit in weights rather than retrieved from external databases, enabling fast inference without retrieval latency.","intents":["I need to get accurate answers to factual questions across multiple domains","I want to understand a complex topic by getting a detailed explanation","I need to verify claims or get context on current events (within training cutoff)","I want to get domain-specific advice (e.g., legal, medical, technical)"],"best_for":["developers building Q&A systems or chatbots","teams creating educational content or tutoring systems","builders prototyping knowledge-intensive applications"],"limitations":["Knowledge cutoff at April 2023; cannot answer questions about events after that date","No real-time information access; cannot browse the web or access live APIs","Prone to hallucination on obscure topics or edge cases not well-represented in training data","Cannot cite sources or provide references for claims; answers are generated without attribution","Performance varies by domain; stronger on well-represented topics (programming, popular science) vs niche areas"],"requires":["OpenAI API key with GPT-4 access","Text input (question or prompt)","No external knowledge base or retrieval system required"],"input_types":["text (natural language questions)","prompts with domain context"],"output_types":["text (answers, explanations, summaries)","structured data (via prompt engineering)"],"categories":["text-generation-language","memory-knowledge"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"openrouter-openai-gpt-4__cap_5","uri":"capability://text.generation.language.instruction.following.with.complex.task.decomposition","name":"instruction-following with complex task decomposition","description":"GPT-4 follows complex, multi-step instructions by decomposing tasks into subtasks and executing them sequentially. Through instruction-tuning on datasets where complex instructions are paired with correct outputs, the model learns to parse task specifications, identify dependencies, and generate outputs that satisfy all constraints. This enables it to handle nuanced requests like 'write a poem in the style of Shakespeare about machine learning, exactly 14 lines, with AABB rhyme scheme'.","intents":["I need the model to follow specific formatting or style constraints","I want to give complex, multi-part instructions and have them all executed","I need to enforce constraints (e.g., word count, tone, structure) on generated content","I want to build systems where users can specify detailed requirements in natural language"],"best_for":["developers building content generation systems with strict requirements","teams creating customizable writing assistants","builders enabling non-technical users to specify complex tasks"],"limitations":["Instruction-following quality degrades with instruction complexity; >5 constraints may be partially ignored","No formal verification that outputs satisfy all constraints; requires post-generation validation","Conflicting instructions may cause the model to prioritize some constraints over others unpredictably","Performance varies based on instruction clarity; ambiguous or poorly-phrased instructions lead to lower compliance"],"requires":["OpenAI API key with GPT-4 access","Clear, well-structured instructions (preferably with examples)","Validation logic to verify constraint satisfaction (external to the model)"],"input_types":["text (complex instructions with multiple constraints)","examples (few-shot demonstrations of desired behavior)"],"output_types":["text (content adhering to specified constraints)","structured outputs (with proper formatting)"],"categories":["text-generation-language","planning-reasoning"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"openrouter-openai-gpt-4__cap_6","uri":"capability://text.generation.language.conversational.context.management.with.multi.turn.dialogue","name":"conversational context management with multi-turn dialogue","description":"GPT-4 maintains conversational context across multiple turns by processing the entire conversation history (user messages and prior assistant responses) as input to each new generation. The model uses transformer attention to track references, pronouns, and implicit context from earlier turns, enabling coherent multi-turn conversations where it can refer back to previous statements, correct itself, or build on prior reasoning. Context is managed by the client; the model itself is stateless.","intents":["I need to build a chatbot that remembers previous messages in a conversation","I want the model to refer back to earlier statements and maintain consistency","I need to enable users to ask follow-up questions that depend on prior context","I want to build interactive debugging or tutoring sessions with multi-turn interactions"],"best_for":["developers building chatbot applications and conversational interfaces","teams creating interactive tutoring or customer support systems","builders enabling multi-turn problem-solving workflows"],"limitations":["Context window limits conversation length; GPT-4 standard supports ~8,000 tokens, limiting ~10-20 turns of typical dialogue","No persistent memory across sessions; each conversation starts fresh without access to prior sessions","Context grows linearly with conversation length, increasing latency and cost per turn","Model may lose track of context in very long conversations (>50 turns) due to attention dilution","No built-in summarization of old context; requires external logic to compress conversation history"],"requires":["OpenAI API key with GPT-4 access","Client-side conversation history management (list of messages with roles)","Token counting logic to stay within context window limits"],"input_types":["text (user message)","conversation history (array of prior messages with roles)"],"output_types":["text (assistant response)","structured data (via prompt engineering)"],"categories":["text-generation-language","memory-knowledge"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"openrouter-openai-gpt-4__cap_7","uri":"capability://text.generation.language.creative.writing.and.content.generation.with.style.control","name":"creative writing and content generation with style control","description":"GPT-4 generates creative content (stories, poems, marketing copy, dialogue) by learning patterns from diverse text sources during pretraining and refining them through instruction-tuning on writing tasks. The model can adopt specific writing styles, tones, and genres by conditioning on style descriptors in the prompt (e.g., 'write in the style of Hemingway'). Generation is controlled through temperature and top-p sampling, enabling trade-offs between creativity (high temperature) and consistency (low temperature).","intents":["I need to generate creative content like stories, poems, or scripts","I want to create marketing copy or product descriptions with a specific tone","I need to generate dialogue for characters or conversational content","I want to brainstorm ideas or get creative variations on a concept"],"best_for":["content creators and writers using AI as a brainstorming tool","marketing teams generating copy variations and ad content","game developers creating dialogue and narrative content","educators creating writing prompts and examples"],"limitations":["Generated content may be derivative or contain unintended plagiarism from training data","Style control is heuristic-based (prompt engineering); no guarantee of consistent style across long outputs","Creative quality is subjective and varies based on prompt quality and sampling parameters","No built-in fact-checking; creative content may contain factual errors or hallucinations","Repetition and clichés are common in longer generated content (>500 tokens)"],"requires":["OpenAI API key with GPT-4 access","Detailed style and tone descriptions in the prompt","Temperature and top-p parameters tuned for desired creativity level"],"input_types":["text (creative brief, style description, genre specification)","examples (few-shot demonstrations of desired style)"],"output_types":["text (stories, poems, marketing copy, dialogue)","structured content (with formatting)"],"categories":["text-generation-language"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"openrouter-openai-gpt-4__cap_8","uri":"capability://text.generation.language.translation.and.multilingual.text.generation.across.100.languages","name":"translation and multilingual text generation across 100+ languages","description":"GPT-4 translates text between 100+ languages and generates content in non-English languages by learning multilingual patterns during pretraining on internet text in diverse languages. The model uses shared transformer parameters across languages, enabling transfer learning where knowledge from high-resource languages (English, Mandarin) improves performance on low-resource languages. Translation quality is improved through instruction-tuning on translation pairs and multilingual instruction-following.","intents":["I need to translate content from English to multiple languages","I want to generate content directly in a non-English language","I need to handle multilingual conversations where users switch languages","I want to localize content for specific regions or language variants"],"best_for":["developers building multilingual applications and global platforms","teams localizing content for international markets","builders creating translation services or multilingual chatbots"],"limitations":["Translation quality varies significantly by language pair; high-resource pairs (English-Spanish) are better than low-resource pairs (English-Icelandic)","No domain-specific terminology handling; technical or specialized translations may be inaccurate","Cultural nuances and idioms may be lost or mistranslated","No real-time language detection; language must be specified in the prompt for best results","Performance degrades on code-mixed text (mixing multiple languages in one sentence)"],"requires":["OpenAI API key with GPT-4 access","Source language and target language specified in the prompt","Text input in the source language"],"input_types":["text (content to translate or prompt for generation)","language specification (source and target languages)"],"output_types":["text (translated or generated content in target language)","multilingual content"],"categories":["text-generation-language"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"openrouter-openai-gpt-4__cap_9","uri":"capability://text.generation.language.summarization.with.configurable.length.and.detail.levels","name":"summarization with configurable length and detail levels","description":"GPT-4 summarizes long documents, articles, or conversations by extracting key information and condensing it into shorter text. The model learns summarization patterns through instruction-tuning on document-summary pairs, enabling it to identify salient information, maintain factual accuracy, and adapt summary length based on prompts (e.g., 'summarize in 2 sentences' or 'provide a detailed summary'). Summarization can be extractive (copying key sentences) or abstractive (paraphrasing and synthesizing).","intents":["I need to condense long documents into brief summaries","I want to extract key points from articles or research papers","I need to summarize conversations or meeting transcripts","I want to create summaries at different detail levels (executive summary vs detailed)"],"best_for":["developers building document processing and knowledge management systems","teams automating meeting notes and transcript summarization","builders creating research tools or content curation platforms"],"limitations":["Summarization quality depends on input clarity; poorly-written or ambiguous documents produce poor summaries","No guarantee of factual accuracy; summaries may contain subtle errors or misinterpretations","Context window limits document length; very long documents (>50,000 tokens) must be chunked","Abstractive summarization may lose specific details or nuance from the original","No built-in citation or reference tracking; summaries don't indicate which parts came from which sections"],"requires":["OpenAI API key with GPT-4 access","Document or text to summarize","Summary length specification (e.g., 'summarize in 100 words')"],"input_types":["text (document, article, conversation transcript)","length specification (word count or detail level)"],"output_types":["text (summary)","structured summaries (with key points extracted)"],"categories":["text-generation-language","data-processing-analysis"],"confidence":0.5,"matches":0,"success_rate":0}],"trust":{"score":25,"verified":false,"data_access_risk":"high","permissions":["OpenAI API key with GPT-4 Vision access enabled","Image input as base64-encoded data or URL (JPEG, PNG, GIF, WebP supported)","HTTP client capable of multipart form submission","OpenAI API key with GPT-4 access","Prompt engineering to trigger reasoning (e.g., 'Let's think step by step')","Sufficient token budget for longer outputs (typically 500-2000 tokens for reasoning traces)","Text to classify","Category definitions (for custom classification)","Optional: few-shot examples for improved accuracy","Unstructured text input"],"failure_modes":["Image resolution capped at ~2000x2000 pixels; larger images are downsampled, losing fine detail","Cannot process video or animated content — only static images","Vision performance degrades on highly stylized or artistic images vs photorealistic content","No real-time video stream processing; requires discrete image submissions","Reasoning quality is prompt-dependent; 'think step by step' is a heuristic, not guaranteed reasoning","No access to intermediate reasoning tokens — only final text output is visible","Reasoning traces can be verbose, increasing token consumption by 2-5x vs direct answers","No formal proof of correctness; reasoning can contain plausible-sounding but incorrect steps","Classification is probabilistic; no confidence scores or uncertainty estimates returned","Performance on custom categories depends on clarity of category definitions and number of examples","builder identity is not verified yet","no observed match outcomes yet"],"rank_breakdown":{"adoption":0.05,"quality":0.5,"ecosystem":0.24,"match_graph":0.25,"freshness":0.75,"weights":{"adoption":0.35,"quality":0.2,"ecosystem":0.1,"match_graph":0.3,"freshness":0.05}},"observed_outcomes":{"matches":0,"success_rate":0,"avg_confidence":0,"top_intents":[],"last_matched_at":null},"maintenance":{"status":"active","updated_at":"2026-05-24T12:16:24.485Z","last_scraped_at":"2026-05-03T15:20:45.777Z","last_commit":null},"community":{"stars":null,"forks":null,"weekly_downloads":null,"model_downloads":null,"model_likes":null}},"distribution":{"claim_url":"https://unfragile.ai/submit?claim=openai-gpt-4","compare_url":"https://unfragile.ai/compare?artifact=openai-gpt-4"}},"signature":"enmTCL+bUxhoLzUvgTf5wsa0/Fd2d0r46zNvmS15DDgi45aZy/cLm9i+YIZNbDtzSeIswXXU5kba9/KjlCu2Cw==","signedAt":"2026-06-22T01:54:17.049Z","signedBy":"unfragile.ai","version":1},"_links":{"self":"https://unfragile.ai/api/v1/passport/openai-gpt-4","artifact":"https://unfragile.ai/openai-gpt-4","verify":"https://unfragile.ai/api/v1/verify?slug=openai-gpt-4","publicKey":"https://unfragile.ai/api/v1/trust-passport-public-key","spec":"https://unfragile.ai/trust","schema":"https://unfragile.ai/schema.json","docs":"https://unfragile.ai/docs"}}