{"passport":{"unfragile":{"@version":"1.0","version":"2026-05","artifact":{"id":"openrouter-morph-morph-v3-large","slug":"morph-morph-v3-large","name":"Morph: Morph V3 Large","type":"model","url":"https://openrouter.ai/models/morph~morph-v3-large","page_url":"https://unfragile.ai/morph-morph-v3-large","categories":["code-editors"],"tags":["morph","api-access","text"],"pricing":{"model":"paid","free":false,"starting_price":"$9.00e-7 per prompt token"},"status":"active","verified":false},"capabilities":[{"id":"openrouter-morph-morph-v3-large__cap_0","uri":"capability://code.generation.editing.structured.code.transformation.with.instruction.guided.ast.manipulation","name":"structured code transformation with instruction-guided ast manipulation","description":"Morph V3 Large accepts code and natural language instructions in a strict XML-like format (<instruction> and <code> tags) and applies precise syntactic and semantic transformations to the code. The model operates on token sequences at ~4,500 tokens/sec, using learned patterns from training data to map instruction semantics to code edits while maintaining syntactic validity. This structured prompt format enables the model to disambiguate instruction intent from code context, reducing hallucination in complex multi-statement edits.","intents":["Apply a specific code refactoring (rename variables, extract functions, change patterns) to a code snippet","Transform code from one style or pattern to another (e.g., callback to async/await, class to functional)","Fix or modify code behavior based on a natural language description of the desired change","Batch-apply the same transformation rule across multiple code samples with high consistency"],"best_for":["Developers building code transformation pipelines or linters that need deterministic, high-accuracy edits","Teams automating large-scale codebase refactoring with validation gates","LLM-powered IDE plugins that apply user-requested code changes with 98%+ correctness"],"limitations":["Requires strict XML-tag formatting of input; malformed prompts will degrade accuracy","98% accuracy means ~1 in 50 transformations may produce syntactically or semantically incorrect code; requires post-generation validation","No built-in context awareness of surrounding codebase; operates only on provided code snippet, limiting cross-file refactoring","Throughput of 4,500 tokens/sec means large batch transformations (>100K tokens) require queuing or parallel requests","No streaming output; full transformed code returned as single response, limiting real-time interactive feedback"],"requires":["OpenRouter API key or direct Morph API access","Prompt formatted with <instruction>{text}</instruction> and <code>{text}</code> XML tags","Code snippet under model's context window (exact limit not specified; assume ~8K-16K tokens typical for LLMs)","HTTP client capable of POST requests with JSON payloads"],"input_types":["code (any programming language)","natural language instruction (English)","structured XML-tagged prompt"],"output_types":["transformed code (same language as input)","text (model may include explanation or confidence markers)"],"categories":["code-generation-editing","code-transformation"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"openrouter-morph-morph-v3-large__cap_1","uri":"capability://code.generation.editing.high.throughput.batch.code.transformation.with.deterministic.output","name":"high-throughput batch code transformation with deterministic output","description":"Morph V3 Large is optimized for throughput at ~4,500 tokens/sec, enabling rapid processing of large batches of code transformation requests. The model produces deterministic outputs for identical inputs (no temperature/sampling randomness in the apply mode), making it suitable for automated pipelines where reproducibility and consistency are critical. The high token-per-second rate allows processing of thousands of code edits in parallel or sequential batches without significant latency accumulation.","intents":["Process hundreds or thousands of code transformation requests in a batch job (e.g., migrate codebase to new API)","Run automated code refactoring across a large repository with consistent, reproducible results","Integrate code transformation into CI/CD pipelines where latency and throughput directly impact build times","Validate code transformation accuracy across test suites by comparing before/after outputs deterministically"],"best_for":["DevOps and platform teams automating large-scale codebase migrations","Linting and code-quality tools that need to apply fixes to thousands of files","Batch processing systems where throughput and cost-per-transformation are optimization targets"],"limitations":["Deterministic output means no creative variation; identical prompts always produce identical code, limiting use cases requiring diverse solutions","4,500 tokens/sec throughput is per-request; parallel requests require managing concurrent API connections and rate limits","No built-in batching API; must orchestrate multiple sequential or parallel requests externally","High throughput assumes small-to-medium code snippets; very large files (>4K tokens) will consume proportionally more time"],"requires":["OpenRouter API key with sufficient rate-limit quota for batch requests","Orchestration layer (e.g., Python script, job queue) to manage multiple sequential/parallel requests","Monitoring of API usage and costs, as batch processing can accumulate significant token spend"],"input_types":["code (multiple snippets in sequence)","natural language instruction (same instruction applied to multiple code samples)"],"output_types":["transformed code (batch results)","structured metadata (e.g., transformation success/failure per snippet)"],"categories":["code-generation-editing","automation-workflow"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"openrouter-morph-morph-v3-large__cap_2","uri":"capability://code.generation.editing.language.agnostic.code.transformation.with.syntax.preservation","name":"language-agnostic code transformation with syntax preservation","description":"Morph V3 Large accepts code in any programming language and applies transformations while preserving syntactic validity. The model learns language-specific patterns during training and applies them at inference time, without requiring explicit language detection or language-specific prompting. This enables a single model to handle Python, JavaScript, Java, Go, Rust, and other languages with consistent accuracy, suggesting the model was trained on diverse language corpora and learned generalizable code transformation patterns.","intents":["Apply the same refactoring instruction (e.g., 'rename variable x to y') across code in different programming languages","Transform code without needing to specify the language or provide language-specific context","Build language-agnostic code transformation tools that work across polyglot codebases"],"best_for":["Teams with polyglot codebases (multiple languages) who need unified code transformation tooling","Language-agnostic linting or refactoring tools that should work across any language","Educational platforms teaching code transformation concepts across multiple languages"],"limitations":["No explicit language detection; if code is ambiguous (e.g., valid in multiple languages), the model may misinterpret intent","Accuracy may vary by language; less common languages (e.g., Kotlin, Elixir) may have lower accuracy than mainstream languages (Python, JavaScript)","No language-specific optimizations; a transformation that's trivial in one language may be complex in another, but the model treats all equally","Syntax preservation is best-effort; edge cases in language-specific syntax (e.g., macros, generics) may not be handled correctly"],"requires":["Code in a language the model was trained on (likely includes Python, JavaScript, Java, Go, Rust, C++, TypeScript, etc.)","No explicit language specification in the prompt; the model infers language from code syntax"],"input_types":["code (any programming language)"],"output_types":["transformed code (same language as input)"],"categories":["code-generation-editing"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"openrouter-morph-morph-v3-large__cap_3","uri":"capability://code.generation.editing.instruction.following.code.generation.with.structured.prompt.enforcement","name":"instruction-following code generation with structured prompt enforcement","description":"Morph V3 Large enforces a strict prompt structure where instructions and code are separated into XML-like tags. This architectural constraint forces the model to learn a clear separation between intent (instruction) and context (code), reducing ambiguity and improving instruction-following accuracy. The model is trained to parse this structure and apply transformations based on the instruction tag, ignoring noise or conflicting signals in the code tag.","intents":["Ensure the model interprets the instruction correctly by structurally separating it from code context","Reduce hallucination or misinterpretation by enforcing a clear input format","Build reliable code transformation APIs where input format is predictable and validated"],"best_for":["Developers building production code transformation services that require high reliability","Teams that want to enforce strict input validation before calling the model","Systems where prompt injection or ambiguous instructions are a concern"],"limitations":["Strict format requirement means any deviation (missing tags, malformed XML) will degrade or break functionality","No flexibility for free-form prompting; users must adapt their workflow to the XML-tag format","Prompt engineering is limited to instruction text; cannot use advanced techniques like few-shot examples or chain-of-thought within the structured format"],"requires":["Prompt formatted exactly as <instruction>{text}</instruction> <code>{text}</code>","Input validation layer to ensure XML tags are present and well-formed before sending to the model"],"input_types":["structured XML-tagged prompt"],"output_types":["transformed code"],"categories":["code-generation-editing"],"confidence":0.5,"matches":0,"success_rate":0}],"trust":{"score":23,"verified":false,"data_access_risk":"low","permissions":["OpenRouter API key or direct Morph API access","Prompt formatted with <instruction>{text}</instruction> and <code>{text}</code> XML tags","Code snippet under model's context window (exact limit not specified; assume ~8K-16K tokens typical for LLMs)","HTTP client capable of POST requests with JSON payloads","OpenRouter API key with sufficient rate-limit quota for batch requests","Orchestration layer (e.g., Python script, job queue) to manage multiple sequential/parallel requests","Monitoring of API usage and costs, as batch processing can accumulate significant token spend","Code in a language the model was trained on (likely includes Python, JavaScript, Java, Go, Rust, C++, TypeScript, etc.)","No explicit language specification in the prompt; the model infers language from code syntax","Prompt formatted exactly as <instruction>{text}</instruction> <code>{text}</code>"],"failure_modes":["Requires strict XML-tag formatting of input; malformed prompts will degrade accuracy","98% accuracy means ~1 in 50 transformations may produce syntactically or semantically incorrect code; requires post-generation validation","No built-in context awareness of surrounding codebase; operates only on provided code snippet, limiting cross-file refactoring","Throughput of 4,500 tokens/sec means large batch transformations (>100K tokens) require queuing or parallel requests","No streaming output; full transformed code returned as single response, limiting real-time interactive feedback","Deterministic output means no creative variation; identical prompts always produce identical code, limiting use cases requiring diverse solutions","4,500 tokens/sec throughput is per-request; parallel requests require managing concurrent API connections and rate limits","No built-in batching API; must orchestrate multiple sequential or parallel requests externally","High throughput assumes small-to-medium code snippets; very large files (>4K tokens) will consume proportionally more time","No explicit language detection; if code is ambiguous (e.g., valid in multiple languages), the model may misinterpret intent","builder identity is not verified yet","no observed match outcomes yet"],"rank_breakdown":{"adoption":0.05,"quality":0.33,"ecosystem":0.24,"match_graph":0.25,"freshness":0.9,"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=morph-morph-v3-large","compare_url":"https://unfragile.ai/compare?artifact=morph-morph-v3-large"}},"signature":"NHwFo1OivlgSp20KdUeVUGJ4DJVujy/ydtAhN6l8c+I5QfwjXbpSS8ZvX5k/5bKtIJcAKRFbX/z8kzqCLUy1Bw==","signedAt":"2026-06-15T18:05:17.213Z","signedBy":"unfragile.ai","version":1},"_links":{"self":"https://unfragile.ai/api/v1/passport/morph-morph-v3-large","artifact":"https://unfragile.ai/morph-morph-v3-large","verify":"https://unfragile.ai/api/v1/verify?slug=morph-morph-v3-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"}}