{"passport":{"unfragile":{"@version":"1.0","version":"2026-05","artifact":{"id":"openrouter-inception-mercury-2","slug":"inception-mercury-2","name":"Inception: Mercury 2","type":"model","url":"https://openrouter.ai/models/inception~mercury-2","page_url":"https://unfragile.ai/inception-mercury-2","categories":["llm-apis"],"tags":["inception","api-access","text"],"pricing":{"model":"paid","free":false,"starting_price":"$2.50e-7 per prompt token"},"status":"active","verified":false},"capabilities":[{"id":"openrouter-inception-mercury-2__cap_0","uri":"capability://planning.reasoning.parallel.token.diffusion.reasoning","name":"parallel-token-diffusion-reasoning","description":"Mercury 2 implements reasoning diffusion LLM (dLLM) architecture that generates and refines multiple tokens in parallel rather than sequentially, using iterative refinement loops to improve token quality across the entire output span simultaneously. This approach reduces latency by distributing computation across token positions instead of the traditional left-to-right autoregressive generation pattern, enabling faster reasoning without sacrificing coherence.","intents":["I need reasoning-heavy responses (math, logic, code analysis) but can't wait for sequential token generation latency","I want to reduce time-to-first-token and total generation time for complex reasoning tasks","I need to understand how a model arrives at conclusions through intermediate reasoning steps"],"best_for":["developers building real-time reasoning agents where latency is critical","teams deploying reasoning models in production with strict SLA requirements","researchers exploring diffusion-based LLM architectures"],"limitations":["parallel refinement may produce different reasoning paths than sequential generation, affecting reproducibility","memory overhead during parallel token refinement could be significant for very long outputs","reasoning quality trade-offs not yet fully characterized vs traditional sequential reasoning models"],"requires":["OpenRouter API access with Mercury 2 model availability","HTTP/REST client capability or OpenRouter SDK","understanding of diffusion-based inference patterns"],"input_types":["text prompts","multi-turn conversation context","structured reasoning queries"],"output_types":["text with reasoning traces","structured reasoning steps","code with explanations"],"categories":["planning-reasoning","text-generation-language"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"openrouter-inception-mercury-2__cap_1","uri":"capability://text.generation.language.fast.inference.latency.optimization","name":"fast-inference-latency-optimization","description":"Mercury 2 is architected for extreme speed through diffusion-based parallel generation, achieving substantially lower end-to-end latency compared to traditional autoregressive LLMs. The model optimizes for time-to-completion rather than token-by-token streaming, making it suitable for synchronous request-response patterns where users expect rapid answers to reasoning queries.","intents":["I need sub-second response times for reasoning queries in production APIs","I want to minimize total latency for complex multi-step reasoning without sacrificing quality","I'm building interactive applications where reasoning speed directly impacts user experience"],"best_for":["production API services with strict latency SLAs (< 5 seconds for reasoning)","interactive applications requiring real-time reasoning feedback","cost-sensitive deployments where faster inference reduces compute costs"],"limitations":["streaming token output may not be available or may be less granular than sequential models","latency benefits diminish for very short queries where overhead dominates","parallel refinement requires sufficient GPU memory; may not scale to extremely long outputs"],"requires":["OpenRouter API endpoint access","network latency < 100ms for optimal performance","support for non-streaming or batch response patterns"],"input_types":["text queries","reasoning prompts","code analysis requests"],"output_types":["complete text responses","reasoning traces","structured outputs"],"categories":["text-generation-language","planning-reasoning"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"openrouter-inception-mercury-2__cap_2","uri":"capability://text.generation.language.multi.turn.reasoning.conversation","name":"multi-turn-reasoning-conversation","description":"Mercury 2 maintains conversation context across multiple turns while applying its parallel diffusion reasoning to each new query, enabling coherent multi-step reasoning dialogues where the model can reference previous reasoning steps and build upon prior conclusions. The architecture preserves context windows while applying fast parallel inference to each turn independently.","intents":["I want to have back-and-forth reasoning conversations where the model remembers and builds on previous logic","I need to debug code or solve problems iteratively with a reasoning model that understands prior context","I'm building a reasoning chatbot that maintains coherent reasoning threads across multiple user turns"],"best_for":["conversational AI applications requiring persistent reasoning context","debugging and problem-solving workflows with iterative refinement","educational tools teaching reasoning through multi-turn dialogue"],"limitations":["context window size limits how much prior reasoning can be referenced (typical 128K-200K tokens)","each turn still requires full parallel refinement, so very long conversations may accumulate latency","context compression or summarization not explicitly mentioned; may lose fine-grained reasoning details in long conversations"],"requires":["OpenRouter API with conversation/chat endpoint support","client-side conversation state management","ability to format multi-turn messages in OpenAI chat format or equivalent"],"input_types":["user messages","system prompts","conversation history"],"output_types":["assistant responses","reasoning traces","follow-up suggestions"],"categories":["text-generation-language","planning-reasoning"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"openrouter-inception-mercury-2__cap_3","uri":"capability://code.generation.editing.code.reasoning.and.analysis","name":"code-reasoning-and-analysis","description":"Mercury 2 applies its fast parallel reasoning to code understanding, generation, and analysis tasks, leveraging reasoning capabilities to explain code logic, identify bugs, suggest optimizations, and generate complex code structures. The diffusion-based approach enables rapid code analysis without the latency overhead of traditional reasoning models.","intents":["I need fast code review and bug detection with reasoning about why issues exist","I want to generate complex code with explanations of the reasoning behind design choices","I need to understand and debug existing code through step-by-step reasoning"],"best_for":["developers using AI for code review and debugging in fast-paced environments","IDE integrations requiring sub-second code analysis responses","code generation tools where reasoning about correctness is important"],"limitations":["code reasoning quality depends on model training data; may struggle with very new languages or frameworks","parallel token generation may produce less deterministic code output than sequential models","no explicit mention of syntax validation or compilation checking"],"requires":["OpenRouter API access","code snippets or files as text input","support for code-specific prompt engineering"],"input_types":["code snippets","full source files","code with comments","error messages and stack traces"],"output_types":["analyzed code","generated code","explanations and reasoning","bug reports with reasoning"],"categories":["code-generation-editing","planning-reasoning"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"openrouter-inception-mercury-2__cap_4","uri":"capability://tool.use.integration.openrouter.api.integration","name":"openrouter-api-integration","description":"Mercury 2 is accessed exclusively through OpenRouter's unified API gateway, which provides standardized request/response formatting, model routing, fallback handling, and usage tracking across multiple LLM providers. Integration uses standard HTTP REST endpoints with OpenAI-compatible chat completion format, enabling drop-in compatibility with existing LLM client libraries.","intents":["I want to use Mercury 2 without managing my own inference infrastructure","I need a unified API that can switch between Mercury 2 and other models without code changes","I want to track usage, costs, and performance metrics for Mercury 2 across my application"],"best_for":["developers building LLM applications who want managed inference without infrastructure","teams using multiple models and needing a unified API layer","startups and small teams without GPU infrastructure"],"limitations":["requires internet connectivity and dependency on OpenRouter's availability","API rate limits and quota management add operational complexity","pricing is per-token and may be higher than self-hosted alternatives for high-volume use","no local inference option; all requests go through OpenRouter's servers"],"requires":["OpenRouter API key (obtained from https://openrouter.ai)","HTTP client library (curl, requests, axios, etc.)","network access to api.openrouter.ai","understanding of OpenAI chat completion API format"],"input_types":["JSON-formatted chat messages","system prompts","user queries"],"output_types":["JSON-formatted completions","usage statistics","error responses"],"categories":["tool-use-integration","text-generation-language"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"openrouter-inception-mercury-2__cap_5","uri":"capability://planning.reasoning.mathematical.reasoning.and.problem.solving","name":"mathematical-reasoning-and-problem-solving","description":"Mercury 2's reasoning capabilities are optimized for mathematical problem-solving, including symbolic manipulation, step-by-step calculation, proof generation, and complex mathematical reasoning. The parallel diffusion approach enables rapid mathematical reasoning without the sequential token overhead that makes traditional reasoning models slow for math-heavy tasks.","intents":["I need to solve complex math problems with step-by-step reasoning and verification","I want to generate mathematical proofs and symbolic derivations quickly","I need to verify mathematical correctness and identify errors in calculations"],"best_for":["educational platforms teaching mathematics with AI tutoring","research tools requiring rapid mathematical reasoning and proof generation","applications solving optimization, calculus, or algebra problems"],"limitations":["mathematical reasoning quality depends on training data; may struggle with very advanced or niche mathematics","no explicit symbolic math engine integration (e.g., SymPy, Mathematica); relies on text-based reasoning","parallel token generation may produce non-deterministic mathematical outputs"],"requires":["OpenRouter API access","mathematical problem statements as text input","optional: LaTeX or mathematical notation support in prompts"],"input_types":["mathematical problem statements","equations","proofs to verify","calculation requests"],"output_types":["step-by-step solutions","mathematical proofs","symbolic derivations","verification results"],"categories":["planning-reasoning","text-generation-language"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"openrouter-inception-mercury-2__cap_6","uri":"capability://planning.reasoning.logical.reasoning.and.deduction","name":"logical-reasoning-and-deduction","description":"Mercury 2 supports logical reasoning tasks including deductive reasoning, constraint satisfaction, logical puzzle solving, and inference chains. The parallel diffusion architecture enables rapid logical reasoning by computing multiple reasoning steps simultaneously rather than sequentially, maintaining logical coherence while reducing latency.","intents":["I need to solve logic puzzles and constraint satisfaction problems quickly","I want to perform deductive reasoning with clear logical chains","I need to verify logical consistency and identify contradictions in arguments"],"best_for":["puzzle and game applications requiring fast logical reasoning","knowledge systems performing inference and deduction","applications verifying logical consistency of complex statements"],"limitations":["logical reasoning quality depends on model training; may struggle with very complex formal logic","no explicit integration with formal logic solvers (e.g., SAT, SMT solvers)","parallel refinement may produce non-deterministic logical outputs"],"requires":["OpenRouter API access","logical problem statements as text input","optional: formal logic notation support"],"input_types":["logical problem statements","constraint descriptions","premises for deduction","logical puzzles"],"output_types":["logical conclusions","deductive chains","consistency verification","solution explanations"],"categories":["planning-reasoning","text-generation-language"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"openrouter-inception-mercury-2__cap_7","uri":"capability://planning.reasoning.reasoning.trace.and.explanation.generation","name":"reasoning-trace-and-explanation-generation","description":"Mercury 2 generates explicit reasoning traces and explanations showing intermediate steps in its reasoning process, enabling transparency into how conclusions are reached. The parallel diffusion approach generates these traces efficiently by refining reasoning steps across the output span simultaneously, making reasoning transparency available without significant latency penalty.","intents":["I need to understand how the model arrived at a conclusion through intermediate reasoning steps","I want to audit and verify reasoning quality by examining the reasoning trace","I need to explain AI decisions to users or stakeholders with clear reasoning chains"],"best_for":["applications requiring explainable AI and reasoning transparency","educational tools teaching reasoning through model explanations","compliance and audit scenarios requiring reasoning documentation"],"limitations":["reasoning traces are generated text, not formal logical proofs; may contain informal reasoning","trace quality depends on model training; may not always produce clear or complete reasoning chains","no explicit structured output format for reasoning traces; requires parsing text output"],"requires":["OpenRouter API access","prompts designed to elicit reasoning traces","optional: post-processing to parse and structure reasoning traces"],"input_types":["queries requesting reasoning explanation","prompts with explicit 'show your reasoning' instructions"],"output_types":["text with reasoning traces","step-by-step explanations","intermediate conclusions","reasoning justifications"],"categories":["planning-reasoning","text-generation-language"],"confidence":0.5,"matches":0,"success_rate":0}],"trust":{"score":24,"verified":false,"data_access_risk":"low","permissions":["OpenRouter API access with Mercury 2 model availability","HTTP/REST client capability or OpenRouter SDK","understanding of diffusion-based inference patterns","OpenRouter API endpoint access","network latency < 100ms for optimal performance","support for non-streaming or batch response patterns","OpenRouter API with conversation/chat endpoint support","client-side conversation state management","ability to format multi-turn messages in OpenAI chat format or equivalent","OpenRouter API access"],"failure_modes":["parallel refinement may produce different reasoning paths than sequential generation, affecting reproducibility","memory overhead during parallel token refinement could be significant for very long outputs","reasoning quality trade-offs not yet fully characterized vs traditional sequential reasoning models","streaming token output may not be available or may be less granular than sequential models","latency benefits diminish for very short queries where overhead dominates","parallel refinement requires sufficient GPU memory; may not scale to extremely long outputs","context window size limits how much prior reasoning can be referenced (typical 128K-200K tokens)","each turn still requires full parallel refinement, so very long conversations may accumulate latency","context compression or summarization not explicitly mentioned; may lose fine-grained reasoning details in long conversations","code reasoning quality depends on model training data; may struggle with very new languages or frameworks","builder identity is not verified yet","no observed match outcomes yet"],"rank_breakdown":{"adoption":0.05,"quality":0.41,"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=inception-mercury-2","compare_url":"https://unfragile.ai/compare?artifact=inception-mercury-2"}},"signature":"GeaCZ6gWaKoX0fC+sV4/csJjYJF0yIaKTQZZMhaQNAy5RVKPu+7uGW8MkOkvKtnBmDfOOnkmrgDznoyyBlwlCw==","signedAt":"2026-06-22T01:56:29.821Z","signedBy":"unfragile.ai","version":1},"_links":{"self":"https://unfragile.ai/api/v1/passport/inception-mercury-2","artifact":"https://unfragile.ai/inception-mercury-2","verify":"https://unfragile.ai/api/v1/verify?slug=inception-mercury-2","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"}}