{"passport":{"unfragile":{"@version":"1.0","version":"2026-05","artifact":{"id":"mistral-large","slug":"mistral-large","name":"Mistral Large","type":"model","url":"https://mistral.ai/news/mistral-large-2/","page_url":"https://unfragile.ai/mistral-large","categories":["llm-apis"],"tags":[],"pricing":{"model":"free","free":true,"starting_price":null},"status":"active","verified":false},"capabilities":[{"id":"mistral-large__cap_0","uri":"capability://memory.knowledge.long.context.reasoning.with.128k.token.window","name":"long-context reasoning with 128k token window","description":"Mistral Large processes up to 128,000 tokens in a single context window, enabling analysis of entire codebases, long documents, or multi-turn conversations without context truncation. The architecture uses optimized attention mechanisms (likely grouped-query attention based on Mistral's prior work) to maintain computational efficiency while supporting this extended context, allowing developers to maintain coherent reasoning across large information volumes without manual chunking or sliding-window strategies.","intents":["analyze an entire codebase for refactoring opportunities without splitting into chunks","process long research papers or documentation for comprehensive summarization","maintain conversation history across 50+ turns without losing earlier context","extract structured data from multi-page documents in a single pass"],"best_for":["enterprise teams processing large documents requiring full-context analysis","developers building code analysis agents that need codebase-wide understanding","research teams working with lengthy academic or technical documents"],"limitations":["latency increases non-linearly with context length; 128K tokens may incur 2-3x latency vs 8K context","cost scales linearly with token count — processing full 128K window is expensive for high-volume applications","retrieval quality degrades in middle sections of very long contexts (lost-in-the-middle effect still present)"],"requires":["API access to Mistral Large via mistral.ai or self-hosted deployment","sufficient token budget for large context processing","client library supporting streaming for long responses (Python SDK 0.0.7+, JavaScript SDK 0.0.7+)"],"input_types":["text","code","structured documents (markdown, JSON, XML)"],"output_types":["text","code","structured analysis"],"categories":["memory-knowledge","planning-reasoning"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"mistral-large__cap_1","uri":"capability://tool.use.integration.native.function.calling.with.schema.based.dispatch","name":"native function calling with schema-based dispatch","description":"Mistral Large implements function calling through a schema-based interface where developers define tool signatures in JSON Schema format, and the model outputs structured function calls that can be directly dispatched to registered handlers. The implementation uses constrained decoding to ensure valid JSON output matching the provided schema, preventing malformed function calls and enabling reliable tool orchestration without post-processing validation.","intents":["call external APIs (weather, database queries, payment processors) directly from model reasoning","build multi-step agents that chain function calls across different services","enable structured tool use without manual prompt engineering for function formatting","integrate with existing REST APIs and microservices through standardized function signatures"],"best_for":["developers building LLM agents requiring reliable tool orchestration","teams integrating Mistral into existing microservice architectures","non-technical builders prototyping AI workflows without deep prompt engineering"],"limitations":["function calling adds ~50-100ms latency per tool invocation due to schema validation and dispatch overhead","maximum function signature complexity is limited; deeply nested schemas may cause parsing failures","no built-in retry logic for failed function calls — requires external orchestration layer","tool context is included in token count, reducing available context for reasoning on large tool sets (50+ functions)"],"requires":["JSON Schema definitions for all tools/functions","API key for Mistral Large (via mistral.ai)","client library with function calling support (Python SDK 0.0.7+, JavaScript SDK 0.0.7+)","handler functions or API endpoints to receive dispatched calls"],"input_types":["text query","JSON Schema tool definitions"],"output_types":["structured function calls (JSON)","function results (any JSON-serializable type)"],"categories":["tool-use-integration","planning-reasoning"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"mistral-large__cap_10","uri":"capability://automation.workflow.self.hosted.deployment.for.data.sovereignty.and.custom.fine.tuning","name":"self-hosted deployment for data sovereignty and custom fine-tuning","description":"Mistral Large can be deployed on-premises or in private cloud environments, enabling organizations to maintain data sovereignty and avoid sending sensitive information to external APIs. Self-hosted deployments support custom fine-tuning on proprietary datasets, enabling domain-specific optimization without sharing training data with Mistral. Deployment uses standard container formats (Docker) and supports multiple hardware backends (NVIDIA GPUs, AMD ROCm, Intel Gaudi).","intents":["deploy AI models in regulated industries (healthcare, finance, government) requiring data residency","fine-tune models on proprietary datasets without exposing training data to third parties","reduce API costs for high-volume inference by running models on owned infrastructure","customize model behavior through fine-tuning on domain-specific examples"],"best_for":["enterprise organizations with data sovereignty requirements","regulated industries (healthcare, finance, government) requiring on-premises deployment","teams with large inference volumes where self-hosting reduces per-token costs"],"limitations":["self-hosting requires significant infrastructure investment (GPU servers, networking, monitoring)","operational overhead includes model serving, scaling, monitoring, and security hardening","fine-tuning requires expertise in machine learning and access to quality training data","custom fine-tuning may degrade performance on general tasks if training data is too narrow","support for self-hosted deployments is limited compared to managed API (no guaranteed SLA)"],"requires":["GPU infrastructure (NVIDIA A100/H100 or equivalent) with sufficient VRAM (48GB+ for 123B model)","container orchestration platform (Kubernetes recommended for production)","model weights and deployment artifacts from Mistral","expertise in ML operations and model serving","optional: fine-tuning infrastructure and training data for custom models"],"input_types":["text prompts","training data for fine-tuning (text, code, domain-specific examples)"],"output_types":["text responses","fine-tuned model weights"],"categories":["automation-workflow","tool-use-integration"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"mistral-large__cap_11","uri":"capability://planning.reasoning.competitive.performance.on.reasoning.benchmarks.vs.gpt.4o.and.claude.3.5","name":"competitive performance on reasoning benchmarks vs gpt-4o and claude 3.5","description":"Mistral Large achieves performance competitive with GPT-4o and Claude 3.5 Sonnet on major reasoning benchmarks including MMLU (84.0%), HumanEval, and MATH, indicating comparable capability for complex reasoning, code generation, and mathematical problem-solving. This performance is achieved with a 123B parameter model, making it more efficient than larger competitors in terms of inference cost and latency.","intents":["replace GPT-4o or Claude 3.5 in existing applications while reducing costs","evaluate Mistral as a primary model for reasoning-heavy workloads","benchmark model quality before committing to long-term vendor relationships","optimize cost-performance tradeoff by choosing Mistral over larger competitors"],"best_for":["cost-conscious teams seeking competitive reasoning capability without premium pricing","organizations evaluating multiple models for production deployment","teams building reasoning-heavy applications with tight budget constraints"],"limitations":["benchmark performance does not guarantee equivalent real-world performance on custom tasks","performance may vary on specialized domains not well-represented in benchmarks","some advanced reasoning tasks may still favor GPT-4o or Claude due to larger model size","benchmark results are point-in-time; newer versions of competitors may exceed Mistral's performance"],"requires":["API key for Mistral Large","evaluation framework to test on your specific use cases","benchmarking infrastructure to compare latency and cost vs alternatives"],"input_types":["reasoning tasks","code generation prompts","mathematical problems"],"output_types":["reasoning outputs","code","mathematical solutions"],"categories":["planning-reasoning","code-generation-editing"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"mistral-large__cap_12","uri":"capability://text.generation.language.temperature.and.sampling.parameter.control.for.output.diversity","name":"temperature and sampling parameter control for output diversity","description":"Mistral Large exposes temperature and top-p (nucleus sampling) parameters to control the randomness and diversity of generated outputs. Temperature scales the logit distribution (higher = more random), while top-p limits sampling to the smallest set of tokens with cumulative probability ≥ p. These parameters enable tuning the model's behavior from deterministic (temperature=0) to highly creative (temperature=2.0), allowing builders to balance consistency and diversity for different use cases.","intents":["generate deterministic outputs for factual tasks by setting temperature to 0","create diverse, creative outputs for brainstorming and content generation by increasing temperature","control hallucination risk by using lower temperature for knowledge-based tasks","fine-tune output quality for specific use cases through parameter experimentation"],"best_for":["developers tuning model behavior for specific applications","teams balancing consistency and creativity in generated content","researchers experimenting with sampling strategies"],"limitations":["parameter effects are non-linear and task-dependent; optimal values require experimentation","very high temperature (>1.5) often produces incoherent or nonsensical outputs","temperature=0 may not be truly deterministic due to floating-point precision; use seed parameter for reproducibility","no built-in guidance on parameter selection; requires domain knowledge or trial-and-error"],"requires":["API key for Mistral API","understanding of temperature and top-p semantics"],"input_types":["text (message) + parameters (temperature: 0-2.0, top_p: 0-1.0)"],"output_types":["text (output diversity controlled by parameters)"],"categories":["text-generation-language","parameter-tuning"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"mistral-large__cap_2","uri":"capability://data.processing.analysis.json.mode.with.schema.enforcement","name":"json mode with schema enforcement","description":"Mistral Large can be constrained to output only valid JSON matching a provided schema, using constrained decoding to enforce structural validity at generation time rather than post-processing. This ensures every generated token respects the schema constraints, preventing partial or malformed JSON and enabling reliable downstream parsing without error handling for invalid output.","intents":["extract structured data from unstructured text with guaranteed valid JSON output","generate configuration files, API payloads, or database records in exact required format","build data pipelines that assume 100% valid JSON without fallback parsing logic","create deterministic outputs for downstream systems that cannot tolerate format variations"],"best_for":["data engineering teams building ETL pipelines requiring guaranteed output format","developers building APIs that return structured responses without validation overhead","teams automating document processing with strict schema requirements"],"limitations":["schema enforcement adds ~30-50ms latency per generation due to constrained decoding overhead","very large schemas (1000+ fields) may cause generation slowdown or token limit issues","model may refuse to generate if schema is too restrictive relative to input (e.g., extracting 50 fields from 100-token document)","enum constraints and pattern validation are supported but complex regex patterns may cause generation failures"],"requires":["JSON Schema definition for output structure","API key for Mistral Large","client library with JSON mode support (Python SDK 0.0.7+, JavaScript SDK 0.0.7+)"],"input_types":["text","unstructured data"],"output_types":["JSON matching provided schema"],"categories":["data-processing-analysis","code-generation-editing"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"mistral-large__cap_3","uri":"capability://text.generation.language.multilingual.reasoning.across.10.languages","name":"multilingual reasoning across 10+ languages","description":"Mistral Large is trained on multilingual data and maintains reasoning capability across 10+ languages including English, French, Spanish, German, Italian, Portuguese, Dutch, Russian, Chinese, Japanese, and Arabic. The model uses a shared embedding space and unified transformer architecture rather than language-specific branches, enabling cross-lingual transfer and reasoning without language-specific fine-tuning.","intents":["build chatbots and support systems serving global audiences without separate language models","translate and reason about code or technical content across language barriers","process customer feedback or documents in multiple languages with consistent quality","create multilingual agents that switch between languages based on user input"],"best_for":["global SaaS platforms requiring multilingual AI without separate model deployments","international teams collaborating across language boundaries","customer support organizations handling tickets in multiple languages"],"limitations":["performance varies by language; non-English languages show 5-15% lower accuracy on benchmarks vs English","code generation quality is highest for English; non-English prompts may produce less idiomatic code","some languages (e.g., low-resource languages) have minimal training data and lower quality","language mixing in single prompt may reduce reasoning quality compared to single-language prompts"],"requires":["API key for Mistral Large","client library supporting UTF-8 text input (Python SDK 0.0.7+, JavaScript SDK 0.0.7+)"],"input_types":["text in any supported language","code with comments in any language"],"output_types":["text in requested language","code with comments in requested language"],"categories":["text-generation-language","planning-reasoning"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"mistral-large__cap_4","uri":"capability://text.generation.language.instruction.following.with.custom.system.prompt.format","name":"instruction-following with custom system prompt format","description":"Mistral Large uses a distinct system prompt format optimized for instruction following, where system instructions are formatted as structured directives that the model interprets with higher fidelity than standard text prompts. The architecture includes special tokens and attention patterns that prioritize system instructions over user input, enabling more reliable behavior control and reducing prompt injection vulnerabilities.","intents":["enforce consistent behavior across different user inputs without complex prompt engineering","build specialized agents with role-based instructions (e.g., 'act as a security auditor')","reduce prompt injection attacks by separating system directives from user input","create deterministic outputs for compliance or safety-critical applications"],"best_for":["teams building production AI systems requiring reliable behavior control","compliance-focused organizations needing auditable instruction adherence","developers building specialized agents with consistent role-based behavior"],"limitations":["system prompt format is Mistral-specific; migrating to other models requires prompt rewriting","very long system prompts (5000+ tokens) may reduce reasoning quality on complex tasks","instruction conflicts (e.g., 'be helpful' vs 'refuse all requests') are not automatically resolved","system instructions cannot override model safety guidelines; safety constraints always apply"],"requires":["API key for Mistral Large","understanding of Mistral's system prompt format (documented in API reference)","client library supporting system messages (Python SDK 0.0.7+, JavaScript SDK 0.0.7+)"],"input_types":["system instructions (text)","user queries (text)"],"output_types":["text following system instructions"],"categories":["text-generation-language","planning-reasoning"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"mistral-large__cap_5","uri":"capability://code.generation.editing.code.generation.and.reasoning.for.40.programming.languages","name":"code generation and reasoning for 40+ programming languages","description":"Mistral Large generates syntactically correct and semantically sound code across 40+ programming languages including Python, JavaScript, Java, C++, Go, Rust, SQL, and domain-specific languages. The model uses language-specific tokenization and training data to understand language idioms, standard libraries, and common patterns, enabling generation of production-quality code with proper error handling and best practices.","intents":["generate boilerplate code and scaffolding for new projects across multiple tech stacks","refactor or optimize existing code while maintaining language idioms and conventions","debug code by analyzing error messages and suggesting fixes with explanations","translate code between languages while preserving logic and adapting to target language patterns"],"best_for":["solo developers building prototypes and MVPs across multiple languages","teams automating code generation in CI/CD pipelines","developers learning new languages and needing idiomatic code examples"],"limitations":["code generation quality varies by language; Python and JavaScript are highest quality, while niche languages (Kotlin, Scala) are lower quality","generated code may lack error handling or edge case coverage; requires human review before production use","context window limits prevent generating entire large applications; best for functions/modules up to 500 lines","no built-in testing or validation; generated code requires unit tests and integration testing"],"requires":["API key for Mistral Large","client library supporting code input/output (Python SDK 0.0.7+, JavaScript SDK 0.0.7+)","knowledge of target language syntax for prompt engineering"],"input_types":["natural language descriptions of desired code","existing code for refactoring or debugging","code snippets in one language for translation"],"output_types":["code in target language","explanations of generated code","debugging suggestions"],"categories":["code-generation-editing","planning-reasoning"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"mistral-large__cap_6","uri":"capability://planning.reasoning.mathematical.reasoning.and.symbolic.computation","name":"mathematical reasoning and symbolic computation","description":"Mistral Large demonstrates strong performance on mathematical reasoning tasks (MATH benchmark: 84.0%) through training on mathematical datasets and symbolic reasoning patterns. The model can solve multi-step math problems, verify proofs, and reason about mathematical concepts without external symbolic engines, though it relies on token-based reasoning rather than formal verification.","intents":["solve mathematical problems step-by-step with explanations for educational applications","verify mathematical proofs and identify logical errors in reasoning","generate mathematical content for textbooks, problem sets, or educational platforms","assist with statistical analysis and data interpretation in research contexts"],"best_for":["educational technology platforms requiring math tutoring capabilities","research teams needing mathematical reasoning assistance","developers building STEM learning applications"],"limitations":["symbolic computation is approximate; for formal verification, external tools (Mathematica, Coq) are required","very complex proofs (100+ steps) may exceed reasoning capability or token limits","numerical precision is limited to floating-point accuracy; high-precision arithmetic requires external libraries","performance degrades on novel or cutting-edge mathematical problems not well-represented in training data"],"requires":["API key for Mistral Large","client library supporting text input (Python SDK 0.0.7+, JavaScript SDK 0.0.7+)","optional: external symbolic math library (SymPy, Mathematica) for verification"],"input_types":["mathematical problems (text or LaTeX)","proofs or mathematical reasoning"],"output_types":["step-by-step solutions","mathematical explanations","proof verification results"],"categories":["planning-reasoning","text-generation-language"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"mistral-large__cap_7","uri":"capability://code.generation.editing.humaneval.code.generation.with.high.pass.rate","name":"humaneval code generation with high pass rate","description":"Mistral Large achieves high performance on HumanEval benchmark (a standard for evaluating code generation quality), generating correct implementations for programming problems that require understanding of algorithms, data structures, and edge cases. The model uses in-context learning from problem descriptions to generate syntactically and semantically correct code without external execution or validation.","intents":["generate correct implementations for algorithmic problems in coding interviews or assessments","create test cases and reference implementations for programming challenges","validate code generation quality for automated code generation pipelines","assist developers in implementing complex algorithms with correct edge case handling"],"best_for":["coding interview preparation platforms and assessment tools","developers building automated code generation systems requiring high correctness","educational platforms evaluating student code against reference implementations"],"limitations":["HumanEval performance (high pass rate) does not guarantee correctness on production code with complex requirements","generated code may pass test cases but lack performance optimization or scalability","edge cases not covered in problem description may not be handled correctly","performance degrades on problems requiring domain-specific knowledge (e.g., graphics, networking)"],"requires":["API key for Mistral Large","client library supporting code generation (Python SDK 0.0.7+, JavaScript SDK 0.0.7+)","test cases or problem descriptions in natural language or structured format"],"input_types":["problem descriptions (natural language or structured)","function signatures or constraints"],"output_types":["code implementations","explanations of approach"],"categories":["code-generation-editing","planning-reasoning"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"mistral-large__cap_8","uri":"capability://text.generation.language.mmlu.benchmark.performance.with.broad.knowledge.coverage","name":"mmlu benchmark performance with broad knowledge coverage","description":"Mistral Large achieves 84.0% accuracy on MMLU (Massive Multitask Language Understanding), a comprehensive benchmark covering 57 tasks across STEM, humanities, social sciences, and professional domains. This performance indicates broad factual knowledge and reasoning capability across diverse domains, though knowledge is frozen at training time and may not reflect recent events.","intents":["answer factual questions across diverse domains for Q&A systems and knowledge bases","provide domain-specific explanations and context for educational or professional applications","validate model knowledge coverage before deployment in specialized domains","generate content that requires broad knowledge synthesis across multiple fields"],"best_for":["general-purpose AI assistants requiring broad knowledge coverage","educational platforms providing explanations across STEM and humanities","professional services (legal, medical, financial) requiring domain knowledge (with appropriate disclaimers)"],"limitations":["knowledge cutoff is fixed at training time; no real-time information about recent events, products, or discoveries","factual accuracy varies by domain; performance is lower on specialized professional knowledge requiring recent updates","model may hallucinate or confabulate facts when knowledge is uncertain; requires fact-checking for critical applications","MMLU performance does not guarantee correctness on domain-specific tasks requiring deep expertise"],"requires":["API key for Mistral Large","client library supporting text input (Python SDK 0.0.7+, JavaScript SDK 0.0.7+)","fact-checking mechanisms for critical applications"],"input_types":["factual questions across any domain","requests for domain-specific explanations"],"output_types":["factual answers","domain-specific explanations","reasoning for answers"],"categories":["text-generation-language","planning-reasoning"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"mistral-large__cap_9","uri":"capability://tool.use.integration.api.based.inference.with.streaming.and.batch.processing","name":"api-based inference with streaming and batch processing","description":"Mistral Large is available via REST API supporting both streaming and batch processing modes. Streaming mode returns tokens incrementally as they are generated, enabling real-time response display and lower time-to-first-token latency. Batch processing mode accepts multiple requests and processes them asynchronously, optimizing throughput for non-real-time applications and reducing per-request overhead.","intents":["build real-time chat interfaces with streaming responses for low-latency user experience","process large volumes of documents or requests asynchronously without blocking","integrate Mistral into existing applications via standard REST API without custom infrastructure","optimize costs by batching requests for non-time-sensitive workloads"],"best_for":["web and mobile applications requiring real-time AI responses","data processing pipelines handling large document volumes","teams without infrastructure expertise seeking managed API access"],"limitations":["API latency is higher than local inference; typical response time is 1-5 seconds depending on output length","streaming adds ~50-100ms overhead per token due to network round-trips","batch processing may have variable latency (minutes to hours) depending on queue depth","API rate limits apply; high-volume applications may require enterprise tier with custom limits","data sent to API is subject to Mistral's privacy policy; sensitive data may require self-hosted deployment"],"requires":["API key from mistral.ai","HTTP client library (requests in Python, fetch in JavaScript)","network connectivity to mistral.ai API endpoints","client library with streaming support (Python SDK 0.0.7+, JavaScript SDK 0.0.7+)"],"input_types":["text prompts","messages in conversation format"],"output_types":["text responses (streaming or batch)","token usage statistics"],"categories":["tool-use-integration","automation-workflow"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"mistral-large__headline","uri":"capability://llm.apis.large.language.model.api.for.advanced.reasoning.and.coding.tasks","name":"large language model api for advanced reasoning and coding tasks","description":"Mistral Large is a state-of-the-art 123B parameter language model designed for high-performance reasoning and coding tasks, featuring a 128K context window and support for multiple languages, making it ideal for developers seeking powerful AI capabilities.","intents":["best large language model API","large language model for coding tasks","top AI model for reasoning benchmarks","enterprise LLM with data sovereignty","Mistral Large vs GPT-4o comparison"],"best_for":["enterprise deployments","high-context tasks"],"limitations":[],"requires":[],"input_types":[],"output_types":[],"categories":["llm-apis"],"confidence":0.5,"matches":0,"success_rate":0}],"trust":{"score":74,"verified":false,"data_access_risk":"high","permissions":["API access to Mistral Large via mistral.ai or self-hosted deployment","sufficient token budget for large context processing","client library supporting streaming for long responses (Python SDK 0.0.7+, JavaScript SDK 0.0.7+)","JSON Schema definitions for all tools/functions","API key for Mistral Large (via mistral.ai)","client library with function calling support (Python SDK 0.0.7+, JavaScript SDK 0.0.7+)","handler functions or API endpoints to receive dispatched calls","GPU infrastructure (NVIDIA A100/H100 or equivalent) with sufficient VRAM (48GB+ for 123B model)","container orchestration platform (Kubernetes recommended for production)","model weights and deployment artifacts from Mistral"],"failure_modes":["latency increases non-linearly with context length; 128K tokens may incur 2-3x latency vs 8K context","cost scales linearly with token count — processing full 128K window is expensive for high-volume applications","retrieval quality degrades in middle sections of very long contexts (lost-in-the-middle effect still present)","function calling adds ~50-100ms latency per tool invocation due to schema validation and dispatch overhead","maximum function signature complexity is limited; deeply nested schemas may cause parsing failures","no built-in retry logic for failed function calls — requires external orchestration layer","tool context is included in token count, reducing available context for reasoning on large tool sets (50+ functions)","self-hosting requires significant infrastructure investment (GPU servers, networking, monitoring)","operational overhead includes model serving, scaling, monitoring, and security hardening","fine-tuning requires expertise in machine learning and access to quality training data","builder identity is not verified yet","no observed match outcomes yet"],"rank_breakdown":{"adoption":0.7,"quality":0.9,"ecosystem":0.15000000000000002,"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":"2026-05-24T12:16:23.328Z","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=mistral-large","compare_url":"https://unfragile.ai/compare?artifact=mistral-large"}},"signature":"0kIJiUZ0GxS/Hh0n+oip1Q8QoQQ7USLjt3qgpgKQskBI8cH4b1qpxgWwSkZdgXTQOQ2/dE7ZNzsn47fGIxebDA==","signedAt":"2026-06-21T05:08:11.404Z","signedBy":"unfragile.ai","version":1},"_links":{"self":"https://unfragile.ai/api/v1/passport/mistral-large","artifact":"https://unfragile.ai/mistral-large","verify":"https://unfragile.ai/api/v1/verify?slug=mistral-large","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"}}