{"passport":{"unfragile":{"@version":"1.0","version":"2026-05","artifact":{"id":"gemma-2-2b","slug":"gemma-2-2b","name":"Gemma 2 2B","type":"model","url":"https://ai.google.dev/gemma","page_url":"https://unfragile.ai/gemma-2-2b","categories":["model-training"],"tags":[],"pricing":{"model":"free","free":true,"starting_price":null},"status":"active","verified":false},"capabilities":[{"id":"gemma-2-2b__cap_0","uri":"capability://text.generation.language.lightweight.text.generation.with.transformer.decoder.architecture","name":"lightweight text generation with transformer decoder architecture","description":"Generates natural language text using a 2-billion-parameter decoder-only transformer architecture optimized for efficiency. The model uses standard transformer attention mechanisms scaled down to fit mobile and edge devices while maintaining coherent multi-turn generation. Inference runs locally on-device or via Google's cloud API, supporting streaming responses for real-time applications.","intents":["Generate conversational responses on mobile devices without cloud latency","Build chatbots and text completion features within strict memory budgets","Deploy language models to IoT and embedded systems with minimal power consumption","Prototype LLM applications without expensive GPU infrastructure"],"best_for":["Mobile app developers building on-device AI features","IoT and embedded systems engineers with <4GB RAM constraints","Startups and indie developers prototyping LLM applications with limited compute budgets","Researchers experimenting with efficient model architectures"],"limitations":["Context window size unknown — may be significantly smaller than 7B+ models, limiting multi-document reasoning","No quantized format specifications documented — unclear if GGUF, int8, or other optimizations are available for local deployment","Performance trade-offs vs larger models unquantified — 'strong relative to size' lacks benchmark comparisons against Phi, Mistral 7B, or other 2B alternatives","Text-only modality — no image, audio, or multimodal understanding capabilities","Inference latency metrics not published — actual on-device speed unknown for different hardware targets"],"requires":["Python 3.8+ (for google-generativeai SDK) or JavaScript/Go/Java/C# for alternative SDKs","API key from Google AI Studio for cloud inference, or local runtime environment for on-device deployment","Minimum 2GB RAM for on-device inference (estimated based on parameter count, unconfirmed)","Internet connection for cloud API access, or offline capability if using quantized local weights"],"input_types":["text (prompts, conversation history, instructions)","structured prompts with system messages and few-shot examples"],"output_types":["text (generated completions, responses, summaries)","streaming token sequences for real-time UI updates"],"categories":["text-generation-language","on-device-inference"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"gemma-2-2b__cap_1","uri":"capability://data.processing.analysis.fine.tuning.and.model.adaptation.for.custom.tasks","name":"fine-tuning and model adaptation for custom tasks","description":"Supports supervised fine-tuning on custom datasets to adapt the base 2B model for domain-specific or task-specific applications. Fine-tuning integrates with Google's training infrastructure via the Generative AI API, allowing developers to update model weights on proprietary data without exposing data to Google's servers (for paid tier users). The capability includes parameter-efficient approaches (likely LoRA or similar, unconfirmed) to reduce computational overhead.","intents":["Adapt Gemma 2 2B to domain-specific language (medical, legal, technical documentation)","Fine-tune on proprietary customer data to improve task-specific accuracy","Create specialized chatbots or assistants with custom knowledge and behavior","Reduce inference costs by using a smaller fine-tuned model instead of larger base models"],"best_for":["Enterprise teams with proprietary datasets and data privacy requirements","Researchers experimenting with efficient fine-tuning methods on small models","Product teams building domain-specific AI features (customer support, content moderation, code generation)","Organizations seeking to reduce inference costs through model specialization"],"limitations":["Fine-tuning methodology not documented — unclear if LoRA, QLoRA, full fine-tuning, or other parameter-efficient methods are used","Training time and cost estimates unavailable — pricing for fine-tuning jobs not specified in provided documentation","No published guidance on minimum dataset size, optimal hyperparameters, or convergence criteria","Data privacy claims ('not used to improve products' for paid tier) unverified — actual data handling practices unknown","No documented support for multi-task or continual learning scenarios"],"requires":["Paid tier access to Google Generative AI API (free tier fine-tuning availability unknown)","Structured training dataset in supported format (format specifications not documented)","API key and authentication credentials for Google Cloud","Python SDK (google-generativeai) or REST API client for job submission"],"input_types":["text training examples (prompt-response pairs)","structured datasets in JSON or CSV format (exact schema unknown)"],"output_types":["fine-tuned model weights or adapter weights","updated model accessible via API with custom model ID"],"categories":["data-processing-analysis","model-training"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"gemma-2-2b__cap_10","uri":"capability://data.processing.analysis.structured.output.generation.with.json.schema.validation","name":"structured output generation with json schema validation","description":"Enables generation of structured outputs (JSON, XML, etc.) by constraining the model's response to match a specified schema. The model generates responses that conform to the provided schema, enabling reliable extraction of structured data without post-processing or parsing. This capability is useful for applications requiring consistent, machine-readable outputs.","intents":["Extract structured data from unstructured text (entity extraction, information extraction)","Generate JSON responses for API integrations","Create consistent output formats for downstream processing","Implement form-filling and data collection through conversational interfaces"],"best_for":["Data extraction and information retrieval applications","API integrations requiring structured responses","Automated data collection and form-filling systems","Applications requiring reliable, machine-readable outputs"],"limitations":["Schema specification format not documented — unclear if JSON Schema, OpenAPI, or custom format is used","Schema complexity limits unknown — very complex schemas may exceed model capabilities","Validation enforcement mechanism unclear — no documentation on how schema violations are handled","No documented performance impact of schema constraints on generation speed","Limited to JSON/XML — no support for custom or domain-specific structured formats"],"requires":["API key for Google Generative AI API","Schema definition in supported format (format unspecified in documentation)","Python 3.8+ SDK or equivalent for other languages"],"input_types":["text prompt","schema definition (JSON Schema or equivalent)"],"output_types":["structured response (JSON, XML, etc.) conforming to schema"],"categories":["data-processing-analysis","text-generation-language"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"gemma-2-2b__cap_11","uri":"capability://safety.moderation.safety.and.content.filtering.with.configurable.guardrails","name":"safety and content filtering with configurable guardrails","description":"Implements content filtering and safety mechanisms to prevent generation of harmful, illegal, or inappropriate content. The model includes built-in safety training and filtering, with configurable safety settings (though specific settings are not documented). Responses flagged as unsafe are blocked or filtered before returning to users.","intents":["Prevent generation of harmful or inappropriate content in production applications","Implement content moderation for user-facing applications","Ensure compliance with content policies and legal requirements","Reduce risk of model misuse for harmful purposes"],"best_for":["Production applications serving general audiences","Applications with strict content policies or legal requirements","Organizations prioritizing safety and responsible AI","Applications handling sensitive domains (healthcare, education, finance)"],"limitations":["Safety settings and configuration options not documented — unclear what guardrails are available","No published safety evaluation or red-teaming results","Safety filtering effectiveness unknown — false positive/negative rates not specified","No transparency into safety training methodology or data","Safety mechanisms may be overly conservative, blocking legitimate requests","No documented appeal or override mechanism for incorrectly filtered responses"],"requires":["API key for Google Generative AI API","Understanding of safety settings (if configurable)","Compliance with Google's usage policies"],"input_types":["text prompt (any content)"],"output_types":["filtered response (if safe) or safety error (if blocked)"],"categories":["safety-moderation","text-generation-language"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"gemma-2-2b__cap_12","uri":"capability://tool.use.integration.token.counting.and.cost.estimation.for.api.usage","name":"token counting and cost estimation for api usage","description":"Provides token counting functionality to estimate API costs before making requests. Developers can count tokens in prompts and responses to calculate expected costs based on per-token pricing. This enables budget planning and cost optimization for applications with variable input sizes.","intents":["Estimate API costs before making requests","Implement cost-aware application logic (e.g., reject requests exceeding budget)","Optimize prompts to reduce token usage and costs","Plan infrastructure budgets based on expected usage patterns"],"best_for":["Cost-conscious applications with variable input sizes","Organizations implementing cost controls and budget limits","Applications optimizing for cost efficiency","Teams planning infrastructure budgets"],"limitations":["Token counting methodology not documented — unclear if it matches actual API token counting","Pricing per token not specified in provided documentation — developers must reference separate pricing page","No batch cost estimation — token counting works on individual requests","No cost tracking or usage analytics in API — requires external logging and analysis","Token counting may differ between estimate and actual usage due to implementation details"],"requires":["API key for Google Generative AI API","Python 3.8+ SDK or equivalent for other languages"],"input_types":["text prompt or message history"],"output_types":["token count (integer)","estimated cost (if pricing is known)"],"categories":["tool-use-integration","data-processing-analysis"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"gemma-2-2b__cap_2","uri":"capability://text.generation.language.multi.language.text.generation.with.language.specific.variants","name":"multi-language text generation with language-specific variants","description":"Generates text in multiple languages through the base Gemma 2 2B model, with specialized variants (TranslateGemma for 55 languages, MedGemma for healthcare) available as separate models. The base model's language coverage is undocumented, but the ecosystem approach allows developers to select language-optimized or domain-optimized variants for specific use cases. All variants share the same 2B parameter efficiency and on-device deployment capability.","intents":["Build multilingual chatbots and customer support systems with a single lightweight model","Generate medical or healthcare-specific text in multiple languages","Translate content between 55 languages using a specialized lightweight model","Deploy language-specific applications to resource-constrained devices in non-English markets"],"best_for":["Global product teams serving non-English markets with limited compute budgets","Healthcare and medical AI applications requiring multilingual support","Localization teams building translation and content generation tools","International development teams in emerging markets with bandwidth/power constraints"],"limitations":["Base model language coverage unknown — no documentation of which languages are supported in the standard Gemma 2 2B","Specialized variants (TranslateGemma, MedGemma) are separate models — requires model selection and switching logic in applications","No cross-lingual transfer or zero-shot translation documented for the base model","Language-specific performance metrics unavailable — quality varies by language but no benchmarks provided","No documented support for code-switching or multilingual prompts within a single generation"],"requires":["Selection of appropriate model variant (base Gemma 2 2B, TranslateGemma, or MedGemma) based on use case","API key for Google Generative AI API or local deployment of selected variant","Language-specific prompt engineering or few-shot examples for optimal results","Python 3.8+ SDK or equivalent for other supported languages"],"input_types":["text prompts in target language","language-specific instructions or system messages","few-shot examples in target language"],"output_types":["text generation in specified language","translated or language-specific responses"],"categories":["text-generation-language","search-retrieval"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"gemma-2-2b__cap_3","uri":"capability://tool.use.integration.cloud.hosted.inference.via.rest.api.and.managed.sdks","name":"cloud-hosted inference via rest api and managed sdks","description":"Provides access to Gemma 2 2B through Google's managed cloud infrastructure via REST API and language-specific SDKs (Python, JavaScript, Go, Java, C#). Inference is handled by Google's servers, eliminating local deployment complexity and providing automatic scaling, load balancing, and infrastructure management. The API supports streaming responses for real-time applications and integrates with Google AI Studio for interactive testing.","intents":["Quickly prototype and test Gemma 2 2B without setting up local infrastructure","Build production applications with automatic scaling and managed infrastructure","Integrate Gemma 2 2B into existing applications via REST API or language-specific SDKs","Stream real-time responses to user interfaces for interactive experiences"],"best_for":["Startups and small teams without DevOps infrastructure","Rapid prototyping and MVP development","Applications requiring automatic scaling and high availability","Teams building web and mobile applications with cloud-native architecture"],"limitations":["Cloud inference introduces network latency — unsuitable for ultra-low-latency applications (e.g., real-time gaming, autonomous systems)","API pricing per token — cost scales with usage, potentially expensive for high-volume applications","Data sent to Google's servers — privacy concerns for sensitive applications (mitigated for paid tier but not fully transparent)","Rate limiting and quota constraints — not documented in provided materials","Dependency on Google's infrastructure — no SLA or uptime guarantees specified"],"requires":["Google Cloud account and API key from Google AI Studio","Internet connectivity for all inference requests","Python 3.8+ (for google-generativeai SDK) or equivalent SDK for other languages","Familiarity with REST API or SDK usage patterns"],"input_types":["text prompts","conversation history (multi-turn chat)","system messages and instructions"],"output_types":["text completions","streaming token sequences","structured responses (JSON if prompted)"],"categories":["tool-use-integration","automation-workflow"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"gemma-2-2b__cap_4","uri":"capability://automation.workflow.on.device.inference.with.local.model.deployment","name":"on-device inference with local model deployment","description":"Enables running Gemma 2 2B directly on mobile devices, IoT hardware, and personal computers without cloud connectivity. The model is optimized for resource-constrained environments through its 2B parameter count and likely includes quantization support (though unconfirmed in documentation). Local inference eliminates network latency, reduces privacy concerns, and enables offline operation, making it suitable for edge AI applications.","intents":["Deploy AI features to mobile apps without cloud dependency or latency","Build offline-capable applications for devices without reliable internet","Reduce privacy concerns by keeping user data on-device","Create responsive AI experiences with sub-100ms inference latency"],"best_for":["Mobile app developers (iOS, Android) building on-device AI features","IoT and embedded systems engineers with strict latency or privacy requirements","Organizations in regions with unreliable internet connectivity","Privacy-conscious applications handling sensitive user data"],"limitations":["Quantization formats and options not documented — unclear if GGUF, int8, or other optimizations are available","Inference speed benchmarks unavailable — actual latency on different hardware (iPhone, Android, Raspberry Pi) unknown","Memory footprint unspecified — estimated 2GB+ RAM required but not confirmed","Battery and thermal impact not documented — power consumption on mobile devices unknown","Model update mechanism unclear — no documented process for deploying new model versions to deployed applications","No built-in support for model caching or incremental updates"],"requires":["Local runtime environment (e.g., TensorFlow Lite, Core ML, ONNX Runtime, or equivalent)","Quantized model weights in supported format (GGUF, safetensors, ONNX, etc. — format availability unknown)","Minimum 2GB RAM and sufficient storage for model weights (estimated)","Mobile SDK or framework integration (e.g., TensorFlow Lite for Android, Core ML for iOS)"],"input_types":["text prompts","user input from mobile UI"],"output_types":["text completions","streaming token sequences for progressive UI updates"],"categories":["automation-workflow","on-device-inference"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"gemma-2-2b__cap_5","uri":"capability://automation.workflow.batch.processing.for.cost.optimized.inference","name":"batch processing for cost-optimized inference","description":"Supports batch API for processing multiple requests asynchronously with 50% cost reduction compared to standard per-token pricing. Batch processing is designed for non-real-time workloads where latency is acceptable in exchange for lower costs. Requests are queued and processed during off-peak hours, making it suitable for bulk content generation, data processing, and analysis tasks.","intents":["Process large volumes of text generation requests at reduced cost","Perform bulk data analysis or content generation overnight","Optimize costs for non-real-time applications like report generation or data labeling","Reduce infrastructure costs for high-volume inference workloads"],"best_for":["Data teams processing large datasets with language models","Content generation platforms with flexible latency requirements","Organizations with high-volume inference needs and cost constraints","Batch processing pipelines and ETL workflows"],"limitations":["Latency is not guaranteed — requests processed asynchronously, potentially taking hours or days","No SLA or completion time guarantees documented","Batch size limits and request queuing behavior not specified","No real-time feedback or progress tracking during batch processing","Unsuitable for interactive or time-sensitive applications"],"requires":["Paid tier access to Google Generative AI API","Batch request format and submission mechanism (details not documented)","Ability to wait for asynchronous processing completion","API key and authentication credentials"],"input_types":["multiple text prompts in batch format","structured batch request payloads"],"output_types":["text completions for all batch requests","batch results file or API response"],"categories":["automation-workflow","data-processing-analysis"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"gemma-2-2b__cap_6","uri":"capability://tool.use.integration.interactive.testing.and.prototyping.via.google.ai.studio","name":"interactive testing and prototyping via google ai studio","description":"Provides a web-based interface (Google AI Studio) for interactive testing, prompt engineering, and experimentation with Gemma 2 2B before deploying to production. The interface supports conversation history, system message configuration, and parameter tuning (temperature, top-k, etc.). Results can be exported as code snippets for integration into applications.","intents":["Experiment with prompts and model behavior without writing code","Tune model parameters (temperature, top-k, top-p) interactively","Generate code snippets for quick integration into applications","Prototype chatbots and conversational AI before production deployment"],"best_for":["Non-technical stakeholders and product managers experimenting with AI capabilities","Developers prototyping and iterating on prompts quickly","Teams evaluating model fit before committing to integration","Rapid MVP development without infrastructure setup"],"limitations":["Limited to interactive testing — not suitable for large-scale evaluation or benchmarking","No built-in evaluation metrics or automated testing framework","No version control or experiment tracking for prompt iterations","Export functionality limited to code snippets — no direct model export or fine-tuning from UI","UI-based limitations on context length and batch processing"],"requires":["Google account and access to Google AI Studio","Web browser with internet connectivity","No coding required for basic testing, but code export requires familiarity with SDKs"],"input_types":["text prompts entered in UI","system messages and instructions","conversation history"],"output_types":["text completions displayed in UI","exportable code snippets (Python, JavaScript, etc.)"],"categories":["tool-use-integration","planning-reasoning"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"gemma-2-2b__cap_7","uri":"capability://text.generation.language.multi.turn.conversation.management.with.context.preservation","name":"multi-turn conversation management with context preservation","description":"Supports multi-turn conversations where the model maintains context across multiple exchanges, enabling natural dialogue and follow-up questions. The API accepts conversation history as a list of messages (user and assistant roles) and generates contextually appropriate responses. Context window size is undocumented, but the model manages conversation state through explicit message passing rather than implicit state management.","intents":["Build conversational chatbots with natural back-and-forth dialogue","Create customer support agents that understand conversation history","Implement interactive tutoring systems with context-aware explanations","Enable follow-up questions and clarifications in dialogue-based applications"],"best_for":["Chatbot and conversational AI applications","Customer support and help desk automation","Interactive tutoring and educational applications","Dialogue-based user interfaces"],"limitations":["Context window size unknown — maximum conversation length before truncation unclear","No documented context management strategy — unclear if oldest messages are dropped or summarized","No built-in conversation persistence — requires external database for storing conversation history","No automatic context summarization or compression for long conversations","Token counting for conversation history not documented — developers must estimate costs"],"requires":["API key for Google Generative AI API","Message history formatted as list of user/assistant exchanges","External storage for persisting conversation history (database, file system, etc.)","Python 3.8+ SDK or equivalent for other languages"],"input_types":["current user message (text)","conversation history (list of user/assistant message pairs)"],"output_types":["assistant response (text)","updated conversation state (for client-side management)"],"categories":["text-generation-language","memory-knowledge"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"gemma-2-2b__cap_8","uri":"capability://text.generation.language.system.message.and.instruction.based.behavior.customization","name":"system message and instruction-based behavior customization","description":"Allows customization of model behavior through system messages and instructions that guide the model's responses without fine-tuning. System messages are prepended to the conversation and establish the model's role, tone, and constraints. This approach enables prompt-based customization for different use cases (customer support, creative writing, technical assistance) without modifying model weights.","intents":["Customize model behavior for different personas (customer support agent, creative writer, technical expert)","Enforce constraints and guidelines (tone, format, content restrictions) through prompting","Implement role-based behavior without fine-tuning","Rapidly iterate on model behavior for different applications"],"best_for":["Teams building multiple applications with different model behaviors","Rapid prototyping and iteration on model behavior","Applications requiring role-based or persona-based responses","Organizations avoiding fine-tuning due to cost or complexity"],"limitations":["System message effectiveness depends on prompt engineering skill — no guarantees on instruction following","No formal specification for system message format or best practices documented","Instruction following quality varies — model may ignore or partially follow system messages","No built-in validation or testing framework for system message effectiveness","System messages consume tokens — longer instructions increase inference costs"],"requires":["Prompt engineering expertise or iterative testing","API key for Google Generative AI API","Understanding of model capabilities and limitations for effective instruction design"],"input_types":["system message (text instruction)","user prompt (text)"],"output_types":["customized response following system message guidance"],"categories":["text-generation-language","planning-reasoning"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"gemma-2-2b__cap_9","uri":"capability://text.generation.language.streaming.response.generation.for.real.time.ui.updates","name":"streaming response generation for real-time ui updates","description":"Supports streaming responses where tokens are returned incrementally as they are generated, enabling real-time UI updates and progressive text display. Streaming reduces perceived latency by showing partial results immediately rather than waiting for the complete response. The API returns token-by-token updates that can be rendered in real-time to users.","intents":["Display text generation results progressively in user interfaces","Reduce perceived latency by showing partial results immediately","Build interactive applications with real-time feedback","Implement typewriter-style text display in chat interfaces"],"best_for":["Web and mobile applications with real-time UI requirements","Chat interfaces and conversational applications","Content generation tools with progressive display","Applications prioritizing user experience and perceived responsiveness"],"limitations":["Streaming adds complexity to client-side implementation — requires handling partial responses and stream termination","Network latency still applies to first token — streaming does not reduce time-to-first-token significantly","Error handling more complex — errors may occur mid-stream, requiring graceful degradation","Streaming not suitable for applications requiring complete response before processing (e.g., structured data extraction)","Token counting for cost estimation more complex with streaming"],"requires":["API key for Google Generative AI API","Client-side streaming support (SDK or HTTP streaming)","UI framework capable of handling incremental updates (React, Vue, etc.)","Python 3.8+ SDK or equivalent for other languages"],"input_types":["text prompt","streaming request flag/parameter"],"output_types":["token stream (incremental text updates)","stream termination signal"],"categories":["text-generation-language","automation-workflow"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"gemma-2-2b__headline","uri":"capability://model.training.lightweight.open.model.for.on.device.applications","name":"lightweight open model for on-device applications","description":"Gemma 2 2B is a lightweight open model with 2 billion parameters, designed for strong performance in on-device applications and fine-tuning experiments, making it ideal for resource-constrained environments.","intents":["best lightweight AI model","open model for on-device applications","AI model for fine-tuning experiments","resource-efficient AI model","2 billion parameter model for inference"],"best_for":["on-device applications","fine-tuning experiments"],"limitations":[],"requires":[],"input_types":[],"output_types":[],"categories":["model-training"],"confidence":0.5,"matches":0,"success_rate":0}],"trust":{"score":57,"verified":false,"data_access_risk":"high","permissions":["Python 3.8+ (for google-generativeai SDK) or JavaScript/Go/Java/C# for alternative SDKs","API key from Google AI Studio for cloud inference, or local runtime environment for on-device deployment","Minimum 2GB RAM for on-device inference (estimated based on parameter count, unconfirmed)","Internet connection for cloud API access, or offline capability if using quantized local weights","Paid tier access to Google Generative AI API (free tier fine-tuning availability unknown)","Structured training dataset in supported format (format specifications not documented)","API key and authentication credentials for Google Cloud","Python SDK (google-generativeai) or REST API client for job submission","API key for Google Generative AI API","Schema definition in supported format (format unspecified in documentation)"],"failure_modes":["Context window size unknown — may be significantly smaller than 7B+ models, limiting multi-document reasoning","No quantized format specifications documented — unclear if GGUF, int8, or other optimizations are available for local deployment","Performance trade-offs vs larger models unquantified — 'strong relative to size' lacks benchmark comparisons against Phi, Mistral 7B, or other 2B alternatives","Text-only modality — no image, audio, or multimodal understanding capabilities","Inference latency metrics not published — actual on-device speed unknown for different hardware targets","Fine-tuning methodology not documented — unclear if LoRA, QLoRA, full fine-tuning, or other parameter-efficient methods are used","Training time and cost estimates unavailable — pricing for fine-tuning jobs not specified in provided documentation","No published guidance on minimum dataset size, optimal hyperparameters, or convergence criteria","Data privacy claims ('not used to improve products' for paid tier) unverified — actual data handling practices unknown","No documented support for multi-task or continual learning scenarios","builder identity is not verified yet","no observed match outcomes yet"],"rank_breakdown":{"adoption":0.7,"quality":0.9,"ecosystem":0.3,"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":"2026-05-24T12:16:21.549Z","last_scraped_at":null,"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=gemma-2-2b","compare_url":"https://unfragile.ai/compare?artifact=gemma-2-2b"}},"signature":"KwHNzMJ4ZzZqVBpZdtNYaNX2JZ4UFZOLQSQHeYO+HhK0UKZehBBWomSKLgzEHHf6zHrBsx+W0cysO/IIlD3rCQ==","signedAt":"2026-06-21T23:46:26.816Z","signedBy":"unfragile.ai","version":1},"_links":{"self":"https://unfragile.ai/api/v1/passport/gemma-2-2b","artifact":"https://unfragile.ai/gemma-2-2b","verify":"https://unfragile.ai/api/v1/verify?slug=gemma-2-2b","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"}}