{"passport":{"unfragile":{"@version":"1.0","version":"2026-05","artifact":{"id":"openrouter-nousresearch-hermes-4-405b","slug":"nousresearch-hermes-4-405b","name":"Nous: Hermes 4 405B","type":"model","url":"https://openrouter.ai/models/nousresearch~hermes-4-405b","page_url":"https://unfragile.ai/nousresearch-hermes-4-405b","categories":["chatbots-assistants"],"tags":["nousresearch","api-access","text"],"pricing":{"model":"paid","free":false,"starting_price":"$1.00e-6 per prompt token"},"status":"active","verified":false},"capabilities":[{"id":"openrouter-nousresearch-hermes-4-405b__cap_0","uri":"capability://planning.reasoning.hybrid.reasoning.with.internal.deliberation","name":"hybrid-reasoning-with-internal-deliberation","description":"Hermes 4 implements a hybrid reasoning architecture where the model dynamically chooses between direct response generation and extended internal deliberation modes. The model uses learned routing mechanisms to determine when complex reasoning chains are necessary versus when direct answers suffice, processing deliberation tokens internally before producing final outputs. This approach reduces unnecessary computation for straightforward queries while enabling deep reasoning for complex problems.","intents":["I need a model that can tackle complex multi-step reasoning problems without always incurring the latency cost of extended thinking","I want the model to automatically decide when to use reasoning versus when to respond directly based on query complexity","I need transparent reasoning traces for debugging and understanding model decision-making on hard problems"],"best_for":["AI researchers building reasoning-intensive applications","Teams developing autonomous agents requiring interpretable decision-making","Developers optimizing for latency-sensitive applications with variable complexity queries"],"limitations":["Hybrid routing adds computational overhead compared to pure inference models; exact latency impact depends on deliberation depth selection","Internal reasoning tokens are not exposed to users by default — requires specific API configuration to access deliberation traces","Performance gains from selective reasoning depend on query distribution; uniform hard problems may not benefit from routing overhead"],"requires":["OpenRouter API key or direct model access through compatible inference provider","Support for extended token context (405B model requires substantial GPU memory or cloud inference)","Client implementation supporting streaming or batch processing of reasoning tokens"],"input_types":["text prompts","multi-turn conversation history","structured reasoning instructions (chain-of-thought, step-by-step)"],"output_types":["text responses","reasoning traces (when enabled)","token usage metadata including deliberation token counts"],"categories":["planning-reasoning","text-generation-language"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"openrouter-nousresearch-hermes-4-405b__cap_1","uri":"capability://text.generation.language.long.context.multi.turn.conversation","name":"long-context-multi-turn-conversation","description":"Hermes 4 supports extended context windows enabling multi-turn conversations with deep history retention and coherent reference resolution across hundreds of exchanges. The model maintains semantic understanding of prior conversation threads, enabling it to track evolving context, resolve pronouns and references to earlier statements, and build upon previous reasoning chains without context collapse. This is implemented through Llama-3.1's optimized attention mechanisms and position interpolation techniques.","intents":["I need to maintain a long-running conversation with consistent character and context across 50+ exchanges","I want the model to remember and reference specific details from earlier in the conversation without explicit re-prompting","I need to build complex multi-step workflows where each turn builds on previous reasoning and decisions"],"best_for":["Developers building conversational AI assistants and chatbots","Teams creating interactive tutoring or mentoring systems requiring sustained context","Researchers studying long-horizon dialogue and context management in LLMs"],"limitations":["Context window size, while large, is finite — extremely long conversations (10,000+ turns) will eventually require summarization or context pruning","Attention complexity grows quadratically with context length; latency increases measurably beyond 100K tokens of context","Model may exhibit recency bias or context dilution in very long conversations, requiring explicit context management strategies"],"requires":["OpenRouter API access or compatible inference provider supporting 405B model","Client-side conversation state management to track turn history and format messages correctly","Sufficient API rate limits and token quotas for extended multi-turn sessions"],"input_types":["text messages","conversation history arrays with role/content structure","system prompts defining conversation context and constraints"],"output_types":["text responses","token usage including prompt and completion token counts","conversation metadata (turn count, total tokens used)"],"categories":["text-generation-language","memory-knowledge"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"openrouter-nousresearch-hermes-4-405b__cap_10","uri":"capability://text.generation.language.summarization.and.information.extraction","name":"summarization-and-information-extraction","description":"Hermes 4 summarizes long documents and extracts key information through instruction-tuning on summarization tasks and pretraining on diverse text corpora. The model can generate abstractive summaries that capture main ideas in condensed form, as well as extractive summaries that identify key sentences. It supports multiple summarization styles (bullet points, paragraphs, headlines) and can extract specific information types (entities, dates, relationships) from unstructured text. This is implemented through attention mechanisms that identify salient information and reasoning about information importance.","intents":["I need to summarize long documents, articles, or reports into concise overviews","I want to extract specific information (entities, dates, relationships) from unstructured text","I need to generate summaries in specific formats (bullet points, headlines, paragraphs) for different use cases"],"best_for":["Teams processing large volumes of documents and extracting key information","Developers building document analysis and search systems","Researchers studying summarization and information extraction"],"limitations":["Summarization quality degrades for very long documents (>10K words); model may lose important details or focus on early/late content","Abstractive summaries may contain hallucinated information not present in source text; requires verification against original","Information extraction accuracy depends on information clarity in source text; ambiguous or poorly-written text produces lower-quality extraction"],"requires":["OpenRouter API access or compatible inference provider","Document or text to summarize/analyze","Optional: specification of summary length, format, or information types to extract"],"input_types":["long documents or articles","unstructured text passages","summary length or format specifications","information type specifications for extraction"],"output_types":["abstractive summaries","extractive summaries","bullet-point summaries","extracted entities and relationships","key information in structured format"],"categories":["text-generation-language","data-processing-analysis"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"openrouter-nousresearch-hermes-4-405b__cap_11","uri":"capability://search.retrieval.semantic.similarity.and.relevance.ranking","name":"semantic-similarity-and-relevance-ranking","description":"Hermes 4 assesses semantic similarity between texts and ranks items by relevance to queries through learned representations and attention mechanisms. The model understands semantic relationships beyond keyword matching, enabling it to identify similar documents even when they use different vocabulary. It can rank search results, recommend similar items, or identify duplicate content based on semantic similarity rather than exact matching. This capability is implemented through pretraining on diverse text corpora and instruction-tuning on relevance ranking tasks.","intents":["I need to find documents or content semantically similar to a query or reference text","I want to rank search results or recommendations by relevance to user intent","I need to identify duplicate or near-duplicate content across large document collections"],"best_for":["Developers building search systems and recommendation engines","Teams managing content deduplication and quality assurance","Researchers studying semantic similarity and relevance ranking"],"limitations":["Semantic similarity assessment may be subjective; model's notion of similarity may not align with domain-specific definitions","Ranking quality depends on query clarity; ambiguous queries produce inconsistent relevance rankings","Computational cost scales with collection size; ranking large document collections requires batching or approximate methods"],"requires":["OpenRouter API access or compatible inference provider","Query or reference text for similarity assessment","Document collection or candidate items to rank"],"input_types":["query text or reference document","candidate documents or items to rank","optional: relevance criteria or ranking instructions"],"output_types":["similarity scores or rankings","ranked lists of documents or items","relevance assessments with explanations"],"categories":["search-retrieval","data-processing-analysis"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"openrouter-nousresearch-hermes-4-405b__cap_12","uri":"capability://text.generation.language.conversational.dialogue.with.personality","name":"conversational-dialogue-with-personality","description":"Hermes 4 engages in natural, personality-consistent dialogue through instruction-tuning on conversational datasets and pretraining on diverse dialogue corpora. The model can adopt specified personas, maintain consistent character traits across conversations, and engage in natural back-and-forth exchanges. It understands conversational conventions (turn-taking, topic transitions, politeness) and can adapt communication style to match user preferences. This is implemented through attention mechanisms that track conversation state and instruction-tuning that enables personality specification.","intents":["I need a conversational AI that can maintain a consistent personality or character throughout interactions","I want the model to adapt its communication style and tone to match user preferences or specified personas","I need natural dialogue that follows conversational conventions and feels authentic rather than robotic"],"best_for":["Developers building conversational AI assistants and chatbots","Teams creating interactive entertainment or gaming experiences","Researchers studying dialogue systems and conversational AI"],"limitations":["Personality consistency may drift in very long conversations; requires periodic reinforcement through system prompts","Model may struggle with maintaining character consistency when given conflicting instructions or user challenges","Dialogue quality depends on persona clarity; vague or contradictory personality specifications produce inconsistent behavior"],"requires":["OpenRouter API access or compatible inference provider","Clear persona or personality specification","Conversation history management for multi-turn interactions"],"input_types":["user messages","persona or personality specifications","conversation history","tone or style guidelines"],"output_types":["natural dialogue responses","personality-consistent messages","responses in specified tone or style"],"categories":["text-generation-language","planning-reasoning"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"openrouter-nousresearch-hermes-4-405b__cap_2","uri":"capability://tool.use.integration.function.calling.with.structured.tool.binding","name":"function-calling-with-structured-tool-binding","description":"Hermes 4 implements structured function calling through schema-based tool binding, where developers define tool specifications as JSON schemas and the model learns to emit properly formatted function calls that map to external APIs or local functions. The model understands tool semantics, parameter requirements, and return types, enabling it to compose multi-step tool sequences and handle tool failures gracefully. This is implemented through instruction-tuning on function-calling datasets and constrained decoding to ensure valid JSON output.","intents":["I need the model to call external APIs or local functions in response to user queries without manual prompt engineering","I want the model to chain multiple tool calls together to solve complex tasks (e.g., search → analyze → summarize)","I need guaranteed valid JSON output for function calls to avoid parsing errors in production systems"],"best_for":["Developers building AI agents that interact with external systems and APIs","Teams creating autonomous workflows that require tool composition and error recovery","Builders implementing retrieval-augmented generation (RAG) systems with tool-based document access"],"limitations":["Model may hallucinate function calls for tools not in its training data; requires explicit schema definition and validation","Tool calling accuracy degrades with schema complexity — deeply nested or ambiguous schemas may produce incorrect parameter bindings","No built-in error recovery; failed tool calls require explicit error handling and re-prompting in the application layer"],"requires":["OpenRouter API with function-calling support enabled, or compatible inference provider","JSON schema definitions for all tools the model should access","Client-side tool execution layer to handle function calls and return results to the model"],"input_types":["text queries","tool schema definitions (JSON Schema format)","conversation history with prior tool calls and results"],"output_types":["function call objects with tool name and parameters","text responses interspersed with tool calls","structured data from tool execution results"],"categories":["tool-use-integration","planning-reasoning"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"openrouter-nousresearch-hermes-4-405b__cap_3","uri":"capability://code.generation.editing.code.generation.and.completion","name":"code-generation-and-completion","description":"Hermes 4 generates code across multiple programming languages through large-scale pretraining on diverse code repositories and instruction-tuning on code-specific tasks. The model understands code structure, semantics, and best practices, enabling it to generate syntactically correct, idiomatic code for various tasks including function implementation, refactoring, and bug fixing. It supports both single-file generation and multi-file context awareness, allowing it to generate code that integrates with existing codebases when provided with sufficient context.","intents":["I need the model to write complete, working functions or classes based on natural language specifications","I want to use the model to refactor or optimize existing code while maintaining functionality","I need the model to understand and generate code that integrates with my existing codebase architecture"],"best_for":["Software developers using AI as a coding assistant for implementation tasks","Teams automating code generation for boilerplate, scaffolding, or repetitive patterns","Researchers studying code generation and program synthesis with large language models"],"limitations":["Generated code may contain subtle bugs or security vulnerabilities; requires human review and testing before production use","Performance degrades for very large files (>10K lines) or complex architectural patterns not well-represented in training data","Language support varies; less common languages may produce lower-quality code than mainstream languages like Python, JavaScript, and Java"],"requires":["OpenRouter API access or compatible inference provider","Code context provided as text (file contents, repository structure, or documentation)","Testing and validation infrastructure to verify generated code correctness"],"input_types":["natural language specifications","existing code snippets or files","code comments and docstrings","test cases or usage examples"],"output_types":["code snippets or complete functions","refactored code","code explanations and documentation","test cases"],"categories":["code-generation-editing","text-generation-language"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"openrouter-nousresearch-hermes-4-405b__cap_4","uri":"capability://text.generation.language.instruction.following.and.task.adaptation","name":"instruction-following-and-task-adaptation","description":"Hermes 4 implements robust instruction-following through extensive instruction-tuning on diverse task datasets, enabling it to understand and execute complex, multi-step instructions with high fidelity. The model learns to parse instruction structure, identify task constraints and requirements, and adapt its behavior accordingly. This includes support for role-playing, style adaptation, output format specification, and conditional logic within instructions. The architecture uses attention mechanisms to track instruction context throughout generation.","intents":["I need the model to follow complex, multi-part instructions with specific output format requirements","I want to specify custom behavior, tone, or style and have the model consistently apply it throughout responses","I need the model to handle conditional instructions (e.g., 'if the user asks X, respond with Y format')"],"best_for":["Developers building specialized AI assistants with custom behavior requirements","Teams creating content generation pipelines with strict formatting and style constraints","Researchers studying instruction-following and task generalization in large language models"],"limitations":["Instruction-following quality degrades with instruction complexity; very long or ambiguous instructions may produce inconsistent results","Model may misinterpret conflicting instructions or fail to prioritize constraints correctly without explicit clarification","Style and tone consistency may drift across very long outputs; requires periodic reinforcement in multi-turn contexts"],"requires":["OpenRouter API access or compatible inference provider","Well-structured, clear instructions provided in system prompts or user messages","Understanding of prompt engineering best practices for optimal instruction clarity"],"input_types":["system prompts with task specifications","natural language instructions","structured instruction formats (JSON, YAML)","examples demonstrating desired behavior"],"output_types":["text responses following specified format","structured data (JSON, CSV, etc.) when format is specified","responses in specified tone, style, or role"],"categories":["text-generation-language","planning-reasoning"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"openrouter-nousresearch-hermes-4-405b__cap_5","uri":"capability://text.generation.language.knowledge.synthesis.and.explanation","name":"knowledge-synthesis-and-explanation","description":"Hermes 4 synthesizes knowledge from its training data to generate comprehensive explanations, summaries, and educational content across diverse domains. The model can break down complex concepts into understandable components, provide examples, and adapt explanation depth to audience level. It uses hierarchical reasoning to structure explanations logically and supports multi-perspective analysis of topics. This capability is implemented through pretraining on educational content and instruction-tuning on explanation tasks.","intents":["I need the model to explain complex technical or scientific concepts in accessible language","I want to generate educational content, tutorials, or learning materials on specific topics","I need the model to provide multiple perspectives or interpretations of a concept or issue"],"best_for":["Educators and content creators building learning materials and tutorials","Technical writers documenting complex systems and APIs","Developers building educational AI assistants and tutoring systems"],"limitations":["Explanations may contain factual inaccuracies or outdated information; knowledge cutoff limits currency of information","Model may oversimplify complex topics or miss nuanced distinctions important to domain experts","Explanation quality varies by domain; well-represented domains in training data produce better explanations than niche topics"],"requires":["OpenRouter API access or compatible inference provider","Clear specification of target audience and desired explanation depth","Fact-checking and domain expert review for high-stakes educational content"],"input_types":["topic or concept to explain","target audience specification","desired explanation format or structure","context or background information"],"output_types":["text explanations","structured outlines or learning paths","examples and analogies","multi-perspective analyses"],"categories":["text-generation-language","memory-knowledge"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"openrouter-nousresearch-hermes-4-405b__cap_6","uri":"capability://text.generation.language.creative.writing.and.content.generation","name":"creative-writing-and-content-generation","description":"Hermes 4 generates creative content including stories, poetry, marketing copy, and other narrative forms through pretraining on diverse creative texts and instruction-tuning on creative writing tasks. The model understands narrative structure, character development, tone, and style, enabling it to generate coherent, engaging creative content. It supports style transfer, genre-specific generation, and collaborative writing workflows where the model extends or refines human-written content.","intents":["I need the model to generate creative stories, poetry, or other narrative content in specific genres or styles","I want to use the model to help with marketing copy, social media content, or other commercial writing","I need the model to extend or refine creative content I've started, maintaining consistency with my voice and vision"],"best_for":["Content creators and writers using AI as a creative tool","Marketing teams generating copy and campaign content at scale","Developers building creative writing assistants or collaborative writing platforms"],"limitations":["Generated creative content may lack originality or exhibit patterns from training data; may produce clichéd or derivative work","Maintaining consistent character voice and narrative coherence degrades over very long outputs (>5000 words)","Model may struggle with niche genres or styles underrepresented in training data"],"requires":["OpenRouter API access or compatible inference provider","Clear specification of genre, style, tone, and any character or world-building constraints","Human editorial review and refinement for publication-quality content"],"input_types":["genre and style specifications","character descriptions or world-building details","opening lines or story premises","tone and voice guidelines"],"output_types":["story text or narrative content","poetry in specified forms","marketing copy and promotional content","character descriptions and dialogue"],"categories":["text-generation-language"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"openrouter-nousresearch-hermes-4-405b__cap_7","uri":"capability://text.generation.language.multilingual.translation.and.localization","name":"multilingual-translation-and-localization","description":"Hermes 4 performs translation and localization across multiple language pairs through pretraining on multilingual corpora and instruction-tuning on translation tasks. The model understands cultural context, idiomatic expressions, and domain-specific terminology, enabling it to produce natural, contextually appropriate translations rather than literal word-for-word conversions. It supports both direct translation and localization tasks that require cultural adaptation beyond simple translation.","intents":["I need to translate content between multiple language pairs while preserving meaning and tone","I want to localize content for specific regions, adapting not just language but cultural references and conventions","I need the model to handle domain-specific terminology and maintain consistency across translated documents"],"best_for":["Content creators and publishers with multilingual audiences","Software teams localizing applications and documentation for international markets","Developers building translation or localization services"],"limitations":["Translation quality varies significantly by language pair; high-resource languages (English, Spanish, French) produce better results than low-resource languages","Model may struggle with very specialized terminology or domain-specific jargon not well-represented in training data","Cultural adaptation requires explicit instruction; model may not automatically localize cultural references without guidance"],"requires":["OpenRouter API access or compatible inference provider","Source language and target language specification","Domain context and terminology glossaries for specialized content","Human review by native speakers for publication-quality translations"],"input_types":["source text in any supported language","target language specification","domain context and terminology glossaries","localization requirements and cultural guidelines"],"output_types":["translated text","localized content with cultural adaptations","terminology consistency reports","translation quality assessments"],"categories":["text-generation-language"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"openrouter-nousresearch-hermes-4-405b__cap_8","uri":"capability://planning.reasoning.question.answering.with.reasoning","name":"question-answering-with-reasoning","description":"Hermes 4 answers questions by retrieving relevant knowledge from its training data and applying reasoning to synthesize answers. The model can handle factual questions, analytical questions requiring inference, and open-ended questions requiring synthesis of multiple perspectives. It uses attention mechanisms to identify relevant knowledge and chain-of-thought reasoning to work through complex questions step-by-step. The hybrid reasoning mode enables the model to choose when to apply extended deliberation for difficult questions.","intents":["I need the model to answer factual questions accurately across diverse domains","I want the model to explain its reasoning for answers, showing how it arrived at conclusions","I need the model to handle ambiguous or complex questions that require synthesis of multiple perspectives"],"best_for":["Developers building question-answering systems and search interfaces","Teams creating knowledge bases and FAQ systems","Researchers studying question-answering and reasoning in large language models"],"limitations":["Knowledge cutoff limits currency of information; model cannot answer questions about events after training data cutoff","Factual accuracy is not guaranteed; model may hallucinate plausible-sounding but incorrect answers, especially for niche topics","Reasoning quality degrades for questions requiring specialized domain knowledge or very recent information"],"requires":["OpenRouter API access or compatible inference provider","Question specification in natural language","Optional: context or background information to improve answer relevance"],"input_types":["natural language questions","context or background information","question type specification (factual, analytical, open-ended)"],"output_types":["text answers","reasoning traces showing how answer was derived","confidence assessments or uncertainty indicators","source citations (when applicable)"],"categories":["planning-reasoning","text-generation-language"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"openrouter-nousresearch-hermes-4-405b__cap_9","uri":"capability://text.generation.language.sentiment.analysis.and.opinion.extraction","name":"sentiment-analysis-and-opinion-extraction","description":"Hermes 4 analyzes sentiment and extracts opinions from text through instruction-tuning on sentiment analysis tasks and pretraining on diverse text corpora. The model can identify sentiment polarity (positive, negative, neutral), intensity, and nuance, as well as extract specific opinions and reasoning behind them. It understands context-dependent sentiment (sarcasm, irony) and can identify sentiment toward specific entities or aspects within text. This is implemented through attention mechanisms that track sentiment-bearing language and reasoning about context.","intents":["I need to analyze sentiment in customer reviews, social media posts, or feedback at scale","I want to extract specific opinions and reasoning from text, not just overall sentiment scores","I need to identify sentiment toward specific entities or aspects within longer text passages"],"best_for":["Teams analyzing customer feedback and reviews","Researchers studying sentiment analysis and opinion mining","Developers building sentiment-aware applications and recommendation systems"],"limitations":["Sentiment analysis accuracy degrades with sarcasm, irony, and context-dependent sentiment; may misclassify intentionally misleading statements","Model may struggle with mixed sentiment (e.g., 'good product but terrible customer service') without explicit instruction to identify multiple sentiments","Language-specific nuances may be missed; sentiment analysis quality varies by language"],"requires":["OpenRouter API access or compatible inference provider","Text input for sentiment analysis","Optional: specification of entities or aspects to analyze sentiment toward"],"input_types":["text passages (reviews, social media posts, feedback)","entity or aspect specifications for targeted sentiment analysis","context or background information"],"output_types":["sentiment polarity (positive, negative, neutral)","sentiment intensity or confidence scores","extracted opinions and reasoning","entity-specific sentiment analysis"],"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":["OpenRouter API key or direct model access through compatible inference provider","Support for extended token context (405B model requires substantial GPU memory or cloud inference)","Client implementation supporting streaming or batch processing of reasoning tokens","OpenRouter API access or compatible inference provider supporting 405B model","Client-side conversation state management to track turn history and format messages correctly","Sufficient API rate limits and token quotas for extended multi-turn sessions","OpenRouter API access or compatible inference provider","Document or text to summarize/analyze","Optional: specification of summary length, format, or information types to extract","Query or reference text for similarity assessment"],"failure_modes":["Hybrid routing adds computational overhead compared to pure inference models; exact latency impact depends on deliberation depth selection","Internal reasoning tokens are not exposed to users by default — requires specific API configuration to access deliberation traces","Performance gains from selective reasoning depend on query distribution; uniform hard problems may not benefit from routing overhead","Context window size, while large, is finite — extremely long conversations (10,000+ turns) will eventually require summarization or context pruning","Attention complexity grows quadratically with context length; latency increases measurably beyond 100K tokens of context","Model may exhibit recency bias or context dilution in very long conversations, requiring explicit context management strategies","Summarization quality degrades for very long documents (>10K words); model may lose important details or focus on early/late content","Abstractive summaries may contain hallucinated information not present in source text; requires verification against original","Information extraction accuracy depends on information clarity in source text; ambiguous or poorly-written text produces lower-quality extraction","Semantic similarity assessment may be subjective; model's notion of similarity may not align with domain-specific definitions","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":"2026-05-24T12:16:24.484Z","last_scraped_at":"2026-05-03T15:20:45.776Z","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=nousresearch-hermes-4-405b","compare_url":"https://unfragile.ai/compare?artifact=nousresearch-hermes-4-405b"}},"signature":"P0HCrHdmDuywMLxjYKeH4sHldUXlwqU9TfTgdixfyz25Gq8C+pJcahBIaS9ogO+YNzwihcloJe6zO4vmdqWwBQ==","signedAt":"2026-06-20T17:44:14.417Z","signedBy":"unfragile.ai","version":1},"_links":{"self":"https://unfragile.ai/api/v1/passport/nousresearch-hermes-4-405b","artifact":"https://unfragile.ai/nousresearch-hermes-4-405b","verify":"https://unfragile.ai/api/v1/verify?slug=nousresearch-hermes-4-405b","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"}}