{ "schema": "x12i.ai-profiles.model-registry.v3", "registryId": "models-profiles", "generatedAt": "2026-05-29T00:00:00.000Z", "lastVerifiedAt": "2026-06-01T12:00:00Z", "version": "4.0.0", "currency": "USD", "pricingUnit": "1M_tokens", "namingConvention": "lowerCamelCase", "semantics": { "provider": "Billing owner / vendor source of truth for the model SKU.", "modelId": "Catalog lookup key (vendor-native id). Direct and OpenRouter slugs live in models-catalog.json.", "openRouterModelId": "Resolved from models-catalog.json at runtime; never stored on profile choices.", "modelStatus": "verified = stable vendor SKU; predicted = forward-looking placeholder — not guaranteed callable.", "requiredCapabilities": "Profile intent contract; concrete model picks remain suggestions until Optimixer selects dynamically.", "deploymentType": "How the model is operated (proprietary API vs self-hosted CPU). Remote profiles require OpenRouter catalog backing.", "cpuLocalProfile": "Only the \"local\" profile runs on-device CPU (funcx / Transformers.js). No OpenRouter.", "inputModalities": "Required input media types for profile intent (text, image, audio, video, file).", "outputModalities": "Required output media types for profile intent (text, image, audio, video, file).", "choiceModalities": "Optional per-choice modality override when catalog linkage is missing.", "profileTags": "Short discovery tags (2–4 chars). Vendor affinity: ds, ggl, xai, mmx, ant, oai. (profile, tag-set) pairs are unique.", "catalogLane": "Required OpenRouter browser lane per profile. resolveAIProfile requires options.catalogLane matching the profile lane.", "secondary": "Optional companion model per choice (review, escalation, verification). Resolved via resolveSecondaryModel; not a failover replacement.", "fallbackChoiceOrder": "Replacement order when constraints exclude the primary choice (cost cap, vendor trim). Independent of secondary.", "volRouting": "Volume agent profile (vol): Flash default, Pro for hard/final. Model SKUs are never profile aliases." }, "sourceCatalog": { "providers": [ { "id": "openai", "docsUrl": "https://platform.openai.com/docs/models" }, { "id": "anthropic", "docsUrl": "https://docs.anthropic.com/en/docs/about-claude/models" }, { "id": "google", "docsUrl": "https://ai.google.dev/gemini-api/docs/models" }, { "id": "deepseek", "docsUrl": "https://api-docs.deepseek.com" }, { "id": "xai", "docsUrl": "https://docs.x.ai/docs/models" }, { "id": "together", "docsUrl": "https://docs.together.ai/docs/serverless-models" }, { "id": "minimax", "docsUrl": "https://platform.minimaxi.com/document" } ], "modelsCatalogUrl": "https://open-assets.x12i.com/models-catalog.json" }, "profiles": { "cheap": { "profile": "cheap", "displayName": "Cheap", "description": "Low-cost model selection for simple, high-volume, non-critical tasks. This does not always mean the absolute cheapest model.", "category": "cost", "tags": [ "eco", "wf", "xtr", "cls", "bat", "tri", "ggl" ], "aliases": [ "economy", "budget", "low-cost", "cost-sensitive" ], "defaultChoice": "default", "choices": { "default": { "choice": "default", "displayName": "Cheap Default", "description": "The recommended cheap model for normal low-cost use.", "provider": "google", "modelId": "gemini-2.5-flash-lite", "runtime": { "instructionTier": "default", "backend": "openrouter", "reasoningEffort": "minimal", "toolPolicy": "none", "outputMode": "text" }, "reason": "Very low-cost model for high-volume extraction, classification, routing, and simple transformations.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:google", "openrouter:catalog" ] }, "modelStatus": "verified", "secondary": { "kind": "choice", "choice": "anthropic_cheap", "reason": "Stronger instruction-following pass when cheap default quality is insufficient." } }, "google_floor": { "choice": "google_floor", "displayName": "Absolute Cheapest", "description": "Use when the main requirement is minimum cost.", "provider": "google", "modelId": "gemini-2.5-flash-lite", "runtime": { "reasoningEffort": "minimal", "executionMode": "batch", "toolPolicy": "none", "outputMode": "text" }, "reason": "Best for background or bulk work where cost matters more than latency or reasoning quality.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:google", "openrouter:catalog" ] }, "modelStatus": "verified" }, "anthropic_cheap": { "choice": "anthropic_cheap", "displayName": "Cheap but Safer", "description": "Use when cheap is required, but you want a stronger instruction-following model.", "provider": "anthropic", "modelId": "claude-haiku-4-5", "runtime": { "reasoningEffort": "low", "toolPolicy": "none", "outputMode": "text" }, "reason": "More expensive than the cost-floor option, but better when instruction following and answer quality matter.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:anthropic", "openrouter:catalog" ] }, "modelStatus": "verified" }, "deepseek_cheap": { "choice": "deepseek_cheap", "displayName": "DeepSeek V4 Flash", "description": "Lowest-cost option for massive volume when even Flash-Lite tiers are too expensive.", "provider": "deepseek", "modelId": "deepseek-v4-flash", "runtime": { "instructionTier": "default", "backend": "openrouter", "reasoningEffort": "minimal", "executionMode": "batch", "toolPolicy": "none", "outputMode": "text" }, "reason": "2026 pricing floor for massive volume tasks where Google's Flash-Lite is still too costly.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:deepseek", "openrouter:catalog" ] }, "modelStatus": "verified" }, "minimax_cheap": { "choice": "minimax_cheap", "displayName": "MiniMax M2.5 Cheap", "description": "Low-cost Chinese-vendor alternative for high-volume non-critical work.", "provider": "minimax", "modelId": "minimax-m2.5", "modelStatus": "verified", "runtime": { "instructionTier": "default", "backend": "openrouter", "reasoningEffort": "minimal", "toolPolicy": "none", "outputMode": "text", "executionMode": "batch" }, "reason": "Competitive APAC pricing with strong coding and office-work fluency for bulk tasks.", "verification": { "lastVerifiedAt": "2026-05-30T00:00:00Z", "sourceRefs": [ "vendor:minimax", "openrouter:catalog" ] } } }, "fallbackChoiceOrder": [ "default", "google_floor", "anthropic_cheap", "deepseek_cheap", "minimax_cheap" ], "requiredCapabilities": { "cost": "low", "reasoning": "minimal", "context": "standard", "inputModalities": [ "text" ], "outputModalities": [ "text" ] }, "catalogLane": "text" }, "value": { "profile": "value", "displayName": "Value", "description": "Cheap but powerful: prioritize strong reasoning per dollar when you want higher capability without flagship pricing.", "category": "cost", "tags": ["eco", "rsn", "cod", "anl", "ds"], "aliases": ["cheap-power", "power-cheap", "value", "cheap-strong"], "defaultChoice": "default", "choices": { "default": { "choice": "default", "displayName": "Value Default (DeepSeek Pro)", "description": "Recommended cheap-but-strong model choice.", "provider": "deepseek", "modelId": "deepseek-v4-pro", "runtime": { "instructionTier": "reasoning", "backend": "openrouter", "reasoningEffort": "high", "toolPolicy": "none", "outputMode": "text" }, "reason": "Strong reasoning-per-dollar alternative when cheap is desired but quality matters." }, "deepseek_flash": { "choice": "deepseek_flash", "displayName": "Value Fast (DeepSeek Flash)", "description": "Cheaper and faster, but still strong for many tasks.", "provider": "deepseek", "modelId": "deepseek-v4-flash", "runtime": { "instructionTier": "default", "backend": "openrouter", "reasoningEffort": "medium", "toolPolicy": "none", "outputMode": "text" }, "reason": "Use when you want the DeepSeek family but need a cheaper/faster option." }, "google_value": { "choice": "google_value", "displayName": "Value Long-Context", "description": "Long-context alternative for heavy reading.", "provider": "google", "modelId": "gemini-3.5-flash", "runtime": { "instructionTier": "default", "backend": "openrouter", "reasoningEffort": "medium", "toolPolicy": "none", "outputMode": "text" }, "reason": "Good alternative when input size and context length matter more than maximum reasoning." } }, "fallbackChoiceOrder": ["default", "deepseek_flash", "google_value"], "requiredCapabilities": { "reasoning": "medium", "coding": true, "cost": "low", "inputModalities": ["text"], "outputModalities": ["text"] }, "catalogLane": "text" }, "testing": { "profile": "testing", "displayName": "Testing", "description": "Cheap default profile for smoke tests, demos, and deterministic integration checks.", "category": "cost", "tags": ["eco", "wf", "bat", "tri"], "aliases": ["test", "testing", "smoke", "smoke-test"], "defaultChoice": "default", "choices": { "default": { "choice": "default", "displayName": "Testing Default", "description": "Stable, cheap baseline model for tests.", "provider": "google", "modelId": "gemini-2.5-flash-lite", "runtime": { "instructionTier": "default", "backend": "openrouter", "reasoningEffort": "minimal", "toolPolicy": "none", "outputMode": "text" }, "reason": "Cheap enough to use in CI-like flows and manual smoke tests." }, "floor": { "choice": "floor", "displayName": "Testing Floor (same SKU)", "description": "Explicit cost-floor choice for tests (intentionally same model).", "provider": "google", "modelId": "gemini-2.5-flash-lite", "runtime": { "reasoningEffort": "minimal", "executionMode": "batch", "toolPolicy": "none", "outputMode": "text" }, "reason": "Same catalog model as default, but labeled separately for easy selection in test matrices." }, "deepseek_test": { "choice": "deepseek_test", "displayName": "Testing DeepSeek", "description": "Alternate vendor for testing provider fallbacks.", "provider": "deepseek", "modelId": "deepseek-v4-flash", "runtime": { "instructionTier": "default", "backend": "openrouter", "reasoningEffort": "low", "toolPolicy": "none", "outputMode": "text" }, "reason": "Useful to test non-Google routing and vendor trimming while staying inexpensive." } }, "fallbackChoiceOrder": ["default", "floor", "deepseek_test"], "requiredCapabilities": { "cost": "low", "reasoning": "minimal", "inputModalities": ["text"], "outputModalities": ["text"] }, "catalogLane": "text" }, "fast": { "profile": "fast", "displayName": "Fast", "description": "Low-latency model selection for interactive product flows.", "category": "speed", "tags": [ "spd", "wf", "agt", "cod", "ggl", "xai" ], "aliases": [ "instant", "quick", "low-latency" ], "defaultChoice": "default", "choices": { "default": { "choice": "default", "displayName": "Fast Default", "description": "Recommended fast model for interactive UX.", "provider": "openai", "modelId": "gpt-4o-mini", "runtime": { "instructionTier": "default", "backend": "openrouter", "reasoningEffort": "low", "toolPolicy": "none", "outputMode": "text" }, "reason": "Good balance of speed, capability, and structured behavior for real-time product usage.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:openai", "openrouter:catalog" ] }, "modelStatus": "verified" }, "google_fast": { "choice": "google_fast", "displayName": "Cheaper Fast", "description": "Fast but more cost-sensitive.", "provider": "google", "modelId": "gemini-3.1-flash-lite", "runtime": { "reasoningEffort": "low", "toolPolicy": "none", "outputMode": "text" }, "reason": "Good when speed and cost are more important than highest reasoning quality.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:google", "openrouter:catalog" ] }, "modelStatus": "verified" }, "anthropic_fast": { "choice": "anthropic_fast", "displayName": "Quality Fast", "description": "Fast, but with stronger quality than the cheapest fast options.", "provider": "anthropic", "modelId": "claude-haiku-4-5", "runtime": { "reasoningEffort": "low", "toolPolicy": "none", "outputMode": "text" }, "reason": "Useful when you want fast Claude-family behavior for short tasks.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:anthropic", "openrouter:catalog" ] }, "modelStatus": "verified" }, "grok_fast": { "choice": "grok_fast", "displayName": "Grok 4.1 Fast", "description": "Ultra-fast interactive model with a very large context window at low cost.", "provider": "xai", "modelId": "grok-build-0.1", "runtime": { "instructionTier": "default", "backend": "openrouter", "reasoningEffort": "low", "toolPolicy": "none", "outputMode": "text" }, "reason": "Strong speed and 2M context at rock-bottom prices for interactive product UX.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:xai", "openrouter:catalog" ] }, "modelStatus": "verified" }, "minimax_fast": { "choice": "minimax_fast", "displayName": "MiniMax M2 Fast", "description": "Compact, low-latency model tuned for agentic and coding workflows.", "provider": "minimax", "modelId": "minimax-m2", "modelStatus": "verified", "runtime": { "instructionTier": "default", "backend": "openrouter", "reasoningEffort": "low", "toolPolicy": "none", "outputMode": "text" }, "reason": "High-efficiency 10B-active model for interactive agent loops at moderate cost.", "verification": { "lastVerifiedAt": "2026-05-30T00:00:00Z", "sourceRefs": [ "vendor:minimax", "openrouter:catalog" ] } } }, "fallbackChoiceOrder": [ "default", "google_fast", "anthropic_fast", "grok_fast", "minimax_fast" ], "requiredCapabilities": { "speed": "high", "cost": "low", "reasoning": "low", "latency": "low", "inputModalities": [ "text" ], "outputModalities": [ "text" ] }, "catalogLane": "text" }, "balanced": { "profile": "balanced", "displayName": "Balanced", "description": "Default model selection for normal graph execution.", "category": "standard", "tags": [ "wf", "anl", "cod", "agt", "ggl" ], "aliases": [ "default", "standard", "normal" ], "defaultChoice": "default", "choices": { "default": { "choice": "default", "displayName": "Balanced Default", "description": "Recommended default for normal AI tasks.", "provider": "openai", "modelId": "gpt-4.1", "runtime": { "instructionTier": "default", "backend": "openrouter", "reasoningEffort": "medium", "toolPolicy": "none", "outputMode": "text" }, "reason": "Good default for general graph execution, coding, product logic, and normal reasoning.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:openai", "openrouter:catalog" ] }, "modelStatus": "verified", "secondary": { "kind": "choice", "choice": "anthropic_balanced", "reason": "Higher-quality reviewer when balanced default needs a second opinion." } }, "anthropic_balanced": { "choice": "anthropic_balanced", "displayName": "Reasoning Balanced", "description": "Balanced profile with stronger reasoning behavior.", "provider": "anthropic", "modelId": "claude-sonnet-4-6", "runtime": { "reasoningEffort": "medium", "toolPolicy": "none", "outputMode": "text" }, "reason": "Good when answer quality, writing quality, and reasoning are more important than raw cost.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:anthropic", "openrouter:catalog" ] }, "modelStatus": "predicted" }, "google_balanced": { "choice": "google_balanced", "displayName": "Cost Balanced", "description": "Balanced but more cost-conscious.", "provider": "google", "modelId": "gemini-3.5-flash", "runtime": { "reasoningEffort": "medium", "toolPolicy": "none", "outputMode": "text" }, "reason": "Good when you want a balanced model but need a lower-cost alternative.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:google", "openrouter:catalog" ] }, "modelStatus": "verified" }, "minimax_balanced": { "choice": "minimax_balanced", "displayName": "MiniMax M2.5 Balanced", "description": "Balanced general workhorse with strong real-world productivity skills.", "provider": "minimax", "modelId": "minimax-m2.5", "modelStatus": "verified", "runtime": { "instructionTier": "default", "backend": "openrouter", "reasoningEffort": "medium", "toolPolicy": "none", "outputMode": "text" }, "reason": "Good mid-tier alternative when you want MiniMax pricing with solid general capability.", "verification": { "lastVerifiedAt": "2026-05-30T00:00:00Z", "sourceRefs": [ "vendor:minimax", "openrouter:catalog" ] } } }, "fallbackChoiceOrder": [ "default", "anthropic_balanced", "google_balanced", "minimax_balanced" ], "requiredCapabilities": { "reasoning": "medium", "coding": true, "context": "standard", "cost": "medium", "inputModalities": [ "text" ], "outputModalities": [ "text" ] }, "catalogLane": "text" }, "deep": { "profile": "deep", "displayName": "Deep", "description": "High-reasoning model selection for complex work.", "category": "reasoning", "tags": [ "rsn", "anl", "cod", "eng", "exp", "pro", "rev", "ds", "xai" ], "aliases": [ "thinking", "reasoning", "high-reasoning", "smart" ], "defaultChoice": "default", "choices": { "default": { "choice": "default", "displayName": "Deep Default", "description": "Recommended high-reasoning model.", "provider": "anthropic", "modelId": "claude-sonnet-4-5", "runtime": { "instructionTier": "reasoning", "backend": "openrouter", "reasoningEffort": "high", "toolPolicy": "none", "outputMode": "text" }, "reason": "Strong default for architecture, planning, code review, analysis, and complex workflows.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:anthropic", "openrouter:catalog" ] }, "modelStatus": "verified", "secondary": { "kind": "choice", "choice": "deepseek_pro", "reason": "Alternate flagship reasoning tier for verification and escalation." } }, "openai_deep": { "choice": "openai_deep", "displayName": "OpenAI Deep", "description": "OpenAI alternative for deep reasoning.", "provider": "openai", "modelId": "gpt-5.4", "runtime": { "reasoningEffort": "high", "toolPolicy": "none", "outputMode": "text" }, "reason": "Strong alternative for coding, architecture, and product reasoning.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:openai", "openrouter:catalog" ] }, "modelStatus": "predicted" }, "google_deep": { "choice": "google_deep", "displayName": "Long-Context Deep", "description": "Deep reasoning with strong long-context handling.", "provider": "google", "modelId": "gemini-3.5-flash", "runtime": { "reasoningEffort": "high", "toolPolicy": "none", "outputMode": "text" }, "reason": "Good alternative when the task needs large-context reading and reasoning.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:google", "openrouter:catalog" ] }, "modelStatus": "verified" }, "deepseek_pro": { "choice": "deepseek_pro", "displayName": "DeepSeek V4 Pro", "description": "High-tier reasoning at a fraction of flagship OpenAI or Anthropic cost.", "provider": "deepseek", "modelId": "deepseek-v4-pro", "runtime": { "instructionTier": "reasoning", "backend": "openrouter", "reasoningEffort": "high", "toolPolicy": "none", "outputMode": "text" }, "reason": "Flagship-class reasoning without Opus or GPT-5.5 pricing.", "verification": { "lastVerifiedAt": "2026-06-01T12:00:00Z", "sourceRefs": [ "vendor:deepseek", "openrouter:catalog", "vendor:deepseek-api-docs" ] }, "modelStatus": "verified" }, "grok_flagship": { "choice": "grok_flagship", "displayName": "Grok 4.3", "description": "Premium alternative reasoning model with a 1M token context window.", "provider": "xai", "modelId": "grok-4.3", "runtime": { "instructionTier": "reasoning", "backend": "openrouter", "reasoningEffort": "high", "toolPolicy": "none", "outputMode": "text" }, "reason": "Alternative premium reasoning with very large context at moderate cost.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:xai", "openrouter:catalog" ] }, "modelStatus": "predicted" }, "minimax_deep": { "choice": "minimax_deep", "displayName": "MiniMax M2.7 Deep", "description": "Flagship MiniMax reasoning with agentic multi-step execution.", "provider": "minimax", "modelId": "minimax-m2.7", "modelStatus": "verified", "runtime": { "instructionTier": "reasoning", "backend": "openrouter", "reasoningEffort": "high", "toolPolicy": "none", "outputMode": "text" }, "reason": "Premium MiniMax tier for complex planning, analysis, and agent workflows.", "verification": { "lastVerifiedAt": "2026-05-30T00:00:00Z", "sourceRefs": [ "vendor:minimax", "openrouter:catalog" ] } } }, "fallbackChoiceOrder": [ "default", "openai_deep", "google_deep", "deepseek_pro", "grok_flagship", "minimax_deep" ], "requiredCapabilities": { "reasoning": "high", "coding": true, "context": "large", "quality": "high", "inputModalities": [ "text" ], "outputModalities": [ "text" ] }, "catalogLane": "text" }, "pro": { "profile": "pro", "displayName": "Pro", "description": "Premium model selection for critical tasks.", "category": "premium", "tags": [ "flg", "pro", "exp", "anl", "cod", "rev" ], "aliases": [ "premium", "best", "highest-quality", "strong" ], "defaultChoice": "default", "choices": { "default": { "choice": "default", "displayName": "Pro Default", "description": "Recommended premium model.", "provider": "openai", "modelId": "gpt-5.5", "runtime": { "instructionTier": "default", "backend": "openrouter", "reasoningEffort": "max", "toolPolicy": "none", "outputMode": "text" }, "reason": "Best for critical reasoning, hard coding, and professional-grade outputs.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:openai", "openrouter:catalog" ] }, "modelStatus": "verified" }, "claude_pro": { "choice": "claude_pro", "displayName": "Claude Pro", "description": "Premium Claude alternative.", "provider": "anthropic", "modelId": "claude-opus-4-8", "runtime": { "reasoningEffort": "max", "toolPolicy": "none", "outputMode": "text" }, "reason": "Best Claude-family option for premium reasoning and writing.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:anthropic", "openrouter:catalog" ] }, "modelStatus": "predicted" }, "anthropic_pro": { "choice": "anthropic_pro", "displayName": "Cost-Aware Pro", "description": "Premium-like behavior with lower cost.", "provider": "anthropic", "modelId": "claude-sonnet-4-6", "runtime": { "reasoningEffort": "high", "toolPolicy": "none", "outputMode": "text" }, "reason": "Good when pro-level quality is desired but Opus/GPT-5.5 cost is too high.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:anthropic", "openrouter:catalog" ] }, "modelStatus": "predicted" } }, "fallbackChoiceOrder": [ "default", "claude_pro", "anthropic_pro" ], "requiredCapabilities": { "reasoning": "high", "quality": "premium", "context": "large", "inputModalities": [ "text" ], "outputModalities": [ "text" ] }, "catalogLane": "text" }, "json": { "profile": "json", "displayName": "JSON", "description": "Model selection for strict JSON, schema, extraction, and classification tasks.", "category": "output", "tags": [ "str", "json", "xtr", "cls", "eco" ], "aliases": [ "schema", "strict-json", "structured", "structured-output" ], "defaultChoice": "default", "runtime": { "outputMode": "schema", "temperature": 0 }, "choices": { "default": { "choice": "default", "displayName": "JSON Default", "description": "Recommended structured-output model.", "provider": "openai", "modelId": "gpt-4o-mini", "runtime": { "instructionTier": "default", "backend": "openrouter", "reasoningEffort": "low", "toolPolicy": "none", "outputMode": "schema", "temperature": 0 }, "reason": "Good default for reliable JSON/schema output at reasonable cost.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:openai", "openrouter:catalog" ] }, "modelStatus": "verified", "secondary": { "kind": "choice", "choice": "anthropic_json", "reason": "Higher-quality structured-output reviewer for complex extraction." } }, "google_json": { "choice": "google_json", "displayName": "Cheap JSON", "description": "Lower-cost structured-output option.", "provider": "google", "modelId": "gemini-3.1-flash-lite", "runtime": { "reasoningEffort": "minimal", "toolPolicy": "none", "outputMode": "schema", "temperature": 0 }, "reason": "Good for high-volume extraction/classification where low cost matters.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:google", "openrouter:catalog" ] }, "modelStatus": "verified" }, "anthropic_json": { "choice": "anthropic_json", "displayName": "Quality JSON", "description": "Structured output with stronger reasoning.", "provider": "anthropic", "modelId": "claude-sonnet-4-6", "runtime": { "reasoningEffort": "medium", "toolPolicy": "none", "outputMode": "schema", "temperature": 0 }, "reason": "Better for complex extraction where reasoning and interpretation are needed.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:anthropic", "openrouter:catalog" ] }, "modelStatus": "predicted" }, "deepseek_json": { "choice": "deepseek_json", "displayName": "DeepSeek V4 Flash JSON", "description": "Low-cost structured output without thinking overhead.", "provider": "deepseek", "modelId": "deepseek-v4-flash", "modelStatus": "verified", "runtime": { "instructionTier": "default", "backend": "openrouter", "reasoningEffort": "minimal", "toolPolicy": "none", "outputMode": "schema", "temperature": 0 }, "reason": "Classification and extraction at DeepSeek Flash pricing; no expensive reasoning.", "verification": { "lastVerifiedAt": "2026-06-01T12:00:00Z", "sourceRefs": [ "vendor:deepseek", "openrouter:catalog" ] } } }, "fallbackChoiceOrder": [ "default", "deepseek_json", "google_json", "anthropic_json" ], "requiredCapabilities": { "structuredOutput": true, "reasoning": "low", "supportsJsonSchema": true, "inputModalities": [ "text" ], "outputModalities": [ "text" ] }, "catalogLane": "text" }, "agentic": { "profile": "agentic", "displayName": "Agentic", "description": "Model selection for tool use, planning, and multi-step execution.", "category": "agentic", "tags": [ "agt", "wf", "cod" ], "aliases": [ "agent", "tools", "tool-use", "tool-calling" ], "defaultChoice": "default", "runtime": { "toolPolicy": "write-with-approval" }, "choices": { "default": { "choice": "default", "displayName": "Agentic Default", "description": "Recommended model for agentic execution.", "provider": "anthropic", "modelId": "claude-sonnet-4-5", "runtime": { "instructionTier": "default", "backend": "openrouter", "reasoningEffort": "high", "toolPolicy": "write-with-approval", "outputMode": "text" }, "reason": "Strong default for multi-step planning, tool use, and agent workflows.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:anthropic", "openrouter:catalog" ] }, "modelStatus": "verified", "secondary": { "kind": "choice", "choice": "openai_agentic", "reason": "Cross-vendor reviewer for agentic tool-use workflows." } }, "openai_agentic": { "choice": "openai_agentic", "displayName": "OpenAI Agentic", "description": "OpenAI alternative for agentic execution.", "provider": "openai", "modelId": "gpt-5.4", "runtime": { "reasoningEffort": "high", "toolPolicy": "write-with-approval", "outputMode": "text" }, "reason": "Good for coding agents, product workflows, and tool orchestration.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:openai", "openrouter:catalog" ] }, "modelStatus": "predicted" }, "google_agentic": { "choice": "google_agentic", "displayName": "Cheap Agentic", "description": "Lower-cost tool-use profile.", "provider": "google", "modelId": "gemini-3.5-flash", "runtime": { "reasoningEffort": "medium", "toolPolicy": "write-with-approval", "outputMode": "text" }, "reason": "Good when the agentic workload is moderate and cost matters.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:google", "openrouter:catalog" ] }, "modelStatus": "verified" }, "minimax_agentic": { "choice": "minimax_agentic", "displayName": "MiniMax M2.7 Agentic", "description": "MiniMax alternative for tool use and multi-step agent execution.", "provider": "minimax", "modelId": "minimax-m2.7", "modelStatus": "verified", "runtime": { "instructionTier": "default", "backend": "openrouter", "reasoningEffort": "high", "toolPolicy": "write-with-approval", "outputMode": "text" }, "reason": "Built for autonomous productivity with strong agentic and planning behavior.", "verification": { "lastVerifiedAt": "2026-05-30T00:00:00Z", "sourceRefs": [ "vendor:minimax", "openrouter:catalog" ] } } }, "fallbackChoiceOrder": [ "default", "deepseek_agentic", "openai_agentic", "google_agentic", "minimax_agentic" ], "requiredCapabilities": { "reasoning": "medium", "coding": true, "supportsTools": true, "context": "large", "inputModalities": [ "text" ], "outputModalities": [ "text" ] }, "catalogLane": "text" }, "research": { "profile": "research", "displayName": "Research", "description": "Model selection for source-heavy reading, synthesis, and cited research.", "category": "research", "tags": [ "res", "anl", "rev" ], "aliases": [ "source-review", "web-research", "deep-research" ], "defaultChoice": "default", "runtime": { "toolPolicy": "read-only", "requireCitations": true, "allowWeb": true, "allowFileSearch": true }, "choices": { "default": { "choice": "default", "displayName": "Research Default", "description": "Recommended research model.", "provider": "anthropic", "modelId": "claude-sonnet-4-5", "runtime": { "instructionTier": "reasoning", "backend": "openrouter", "reasoningEffort": "high", "toolPolicy": "read-only", "outputMode": "text", "requireCitations": true, "allowWeb": true, "allowFileSearch": true }, "reason": "Strong default for reading, comparing sources, and synthesizing long-context information.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:anthropic", "openrouter:catalog" ] }, "modelStatus": "verified" }, "openai_research": { "choice": "openai_research", "displayName": "Premium Research", "description": "Higher-quality research profile.", "provider": "openai", "modelId": "gpt-5.5", "runtime": { "reasoningEffort": "max", "toolPolicy": "read-only", "outputMode": "text", "requireCitations": true, "allowWeb": true, "allowFileSearch": true }, "reason": "Good when research quality matters more than cost.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:openai", "openrouter:catalog" ] }, "modelStatus": "verified" }, "google_research": { "choice": "google_research", "displayName": "Long-Context Research", "description": "Research profile optimized for large context.", "provider": "google", "modelId": "gemini-3.5-flash", "runtime": { "reasoningEffort": "high", "toolPolicy": "read-only", "outputMode": "text", "requireCitations": true, "allowWeb": true, "allowFileSearch": true }, "reason": "Good when research requires very large context windows.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:google", "openrouter:catalog" ] }, "modelStatus": "verified" } }, "fallbackChoiceOrder": [ "default", "openai_research", "google_research" ], "requiredCapabilities": { "reasoning": "high", "context": "large", "quality": "high", "inputModalities": [ "text" ], "outputModalities": [ "text" ] }, "catalogLane": "text" }, "cyber": { "profile": "cyber", "displayName": "Cyber & Risk Analysis", "description": "Model selection tuned for threat analysis, TPRM, contract verification, log parsing, and security research.", "category": "security", "tags": [ "cyb", "sec", "anl", "for", "aud", "exp", "eco", "res", "ds" ], "aliases": [ "security", "threat-intel", "risk", "audit", "cyber-analysis", "cybersecurity" ], "defaultChoice": "default", "choices": { "default": { "choice": "default", "displayName": "Cyber Default (DeepSeek V4 Flash)", "description": "Default cyber analysis: fast, low-cost, 1M context for most TPRM, triage, and agent steps.", "provider": "deepseek", "modelId": "deepseek-v4-flash", "runtime": { "instructionTier": "default", "backend": "openrouter", "reasoningEffort": "medium", "toolPolicy": "read-only", "outputMode": "text" }, "reason": "Best default for high-volume cyber workflows; escalate to Pro or Grok for harder forensics.", "verification": { "lastVerifiedAt": "2026-06-01T12:00:00Z", "sourceRefs": [ "vendor:deepseek", "openrouter:catalog", "vendor:deepseek-api-docs" ] }, "modelStatus": "verified" }, "grok_intel": { "choice": "grok_intel", "displayName": "Grok 4.3 Intel", "description": "Real-time threat intel and large-context log review (1M tokens).", "provider": "xai", "modelId": "grok-4.3", "runtime": { "instructionTier": "reasoning", "backend": "openrouter", "reasoningEffort": "high", "toolPolicy": "read-only", "outputMode": "text" }, "reason": "Strong for bulk log ingestion, MCP workflows, and comparative cyber research vs GPT-5.5.", "verification": { "lastVerifiedAt": "2026-06-01T12:00:00Z", "sourceRefs": [ "vendor:xai", "openrouter:catalog" ] }, "modelStatus": "verified" }, "log_analysis": { "choice": "log_analysis", "displayName": "Bulk Log Parser", "description": "High-volume structured parsing of server logs or network traffic.", "provider": "google", "modelId": "gemini-3.5-flash", "runtime": { "reasoningEffort": "medium", "toolPolicy": "none", "outputMode": "schema" }, "reason": "Massive context for raw security logs and structured JSON alerts at lower cost.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:google", "openrouter:catalog" ] }, "modelStatus": "verified" }, "deep_forensics": { "choice": "deep_forensics", "displayName": "Deep Forensics (DeepSeek V4 Pro)", "description": "Hard reasoning for vulnerability hunting, exploit chaining, and code auditing.", "provider": "deepseek", "modelId": "deepseek-v4-pro", "runtime": { "instructionTier": "reasoning", "backend": "openrouter", "reasoningEffort": "high", "toolPolicy": "read-only", "outputMode": "text" }, "reason": "Economic pro-tier forensics; use gpt_forensics when you need OpenAI Trusted Access validation paths.", "verification": { "lastVerifiedAt": "2026-06-01T12:00:00Z", "sourceRefs": [ "vendor:deepseek", "openrouter:catalog" ] }, "modelStatus": "verified" }, "gpt_forensics": { "choice": "gpt_forensics", "displayName": "GPT-5.5 Forensics (flagship)", "description": "Maximum validated reasoning for zero-day research and complex attack paths.", "provider": "openai", "modelId": "gpt-5.5", "runtime": { "instructionTier": "reasoning", "backend": "openrouter", "reasoningEffort": "max", "toolPolicy": "read-only", "outputMode": "text" }, "reason": "Flagship alternative when DeepSeek Pro is not enough for validated deep exploit chains.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:openai", "openrouter:catalog" ] }, "modelStatus": "verified" } }, "fallbackChoiceOrder": [ "default", "grok_intel", "log_analysis", "deep_forensics", "gpt_forensics" ], "requiredCapabilities": { "reasoning": "high", "coding": true, "context": "large", "inputModalities": [ "text" ], "outputModalities": [ "text" ] }, "catalogLane": "text" }, "content": { "profile": "content", "displayName": "Content Generation", "description": "Model selection optimized for prose, marketing, and programmatic SEO.", "category": "creative", "tags": [ "cre", "wrt", "bat" ], "aliases": [ "writing", "creative", "seo", "copy" ], "defaultChoice": "default", "choices": { "default": { "choice": "default", "displayName": "Prose Default", "description": "Standard high-quality writing.", "provider": "anthropic", "modelId": "claude-sonnet-4-5", "runtime": { "instructionTier": "default", "backend": "openrouter", "reasoningEffort": "medium", "toolPolicy": "none", "outputMode": "text" }, "reason": "Natural, least AI-sounding prose for marketing and long-form copy.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:anthropic", "openrouter:catalog" ] }, "modelStatus": "verified" }, "pseo_bulk": { "choice": "pseo_bulk", "displayName": "Programmatic SEO", "description": "High-speed, low-cost generation for verticalized landing pages.", "provider": "google", "modelId": "gemini-3.1-flash-lite", "runtime": { "reasoningEffort": "low", "executionMode": "batch", "toolPolicy": "none", "outputMode": "schema" }, "reason": "Hundreds of persona-led pages need cost efficiency and structured front-matter output.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:google", "openrouter:catalog" ] }, "modelStatus": "verified" }, "brand_voice": { "choice": "brand_voice", "displayName": "Premium Copy", "description": "Highest-quality creative writing for hero sections and manifestos.", "provider": "anthropic", "modelId": "claude-opus-4-8", "runtime": { "reasoningEffort": "high", "toolPolicy": "none", "outputMode": "text" }, "reason": "When brand voice, tone, and persuasive copywriting are paramount.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:anthropic", "openrouter:catalog" ] }, "modelStatus": "predicted" } }, "fallbackChoiceOrder": [ "default", "pseo_bulk", "brand_voice" ], "requiredCapabilities": { "reasoning": "medium", "context": "standard", "quality": "high", "inputModalities": [ "text" ], "outputModalities": [ "text" ] }, "catalogLane": "text" }, "vision": { "profile": "vision", "displayName": "Vision & Multimodal", "description": "Model selection for image analysis, OCR, and spatial reasoning.", "category": "multimodal", "tags": [ "vis", "anl" ], "aliases": [ "image", "ocr", "visual" ], "defaultChoice": "default", "choices": { "default": { "choice": "default", "displayName": "Vision Default", "description": "General-purpose image understanding.", "provider": "openai", "modelId": "gpt-4.1", "runtime": { "instructionTier": "default", "backend": "openrouter", "reasoningEffort": "medium", "toolPolicy": "none", "outputMode": "text" }, "reason": "Strong spatial reasoning and chart interpretation.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:openai", "openrouter:catalog" ] }, "modelStatus": "verified" }, "document_ocr": { "choice": "document_ocr", "displayName": "Document OCR", "description": "Extracting text and structure from scanned documents and PDFs.", "provider": "anthropic", "modelId": "claude-sonnet-4-6", "runtime": { "reasoningEffort": "medium", "toolPolicy": "none", "outputMode": "schema" }, "reason": "Maintains document hierarchy and outputs extracted data to strict JSON schemas.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:anthropic", "openrouter:catalog" ] }, "modelStatus": "predicted" } }, "fallbackChoiceOrder": [ "default", "document_ocr" ], "requiredCapabilities": { "reasoning": "medium", "supportsVision": true, "context": "large", "inputModalities": [ "text", "image" ], "outputModalities": [ "text" ] }, "catalogLane": "image" }, "local": { "profile": "local", "displayName": "Local / CPU (no OpenRouter)", "description": "On-device inference only — llama.cpp GGUF or Transformers.js on CPU. Does not call OpenRouter or any cloud LLM API.", "category": "local", "tags": [ "loc" ], "aliases": [ "weak", "offline", "on-device" ], "defaultChoice": "llama_cpp_gguf", "deploymentType": "selfHosted", "metadata": { "transport": "cpu-local", "openRouter": "cpu-local", "billingModel": "self_hosted" }, "choices": { "llama_cpp_gguf": { "choice": "llama_cpp_gguf", "displayName": "llama.cpp GGUF (CPU, offline)", "description": "Runs a local GGUF weights file via node-llama-cpp. Configure LLAMA_CPP_MODEL_PATH. No network LLM calls.", "provider": "local", "modelId": "funcx:llama-cpp:gguf", "runtime": { "instructionTier": "cpu", "executionMode": "offline", "backend": "llama-cpp", "reasoningEffort": "minimal", "temperature": 0.1, "toolPolicy": "none", "outputMode": "text" }, "metadata": { "openRouter": "cpu-local", "funcxBackend": "llama-cpp", "envModelPath": "LLAMA_CPP_MODEL_PATH", "envThreads": "LLAMA_CPP_THREADS", "envContextSize": "LLAMA_CPP_CONTEXT_SIZE", "defaultModelPath": "./models/model.gguf", "optionalDep": "node-llama-cpp", "billingModel": "self_hosted", "device": "cpu" }, "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:local", "cpu:self-hosted" ] }, "modelStatus": "verified", "deploymentType": "selfHosted" }, "transformersjs_default": { "choice": "transformersjs_default", "displayName": "Transformers.js (CPU, offline)", "description": "Runs a Hugging Face model in-browser/on-node via @huggingface/transformers on CPU only. No OpenRouter.", "provider": "local", "modelId": "Xenova/distilbart-cnn-6-6", "runtime": { "instructionTier": "cpu", "executionMode": "offline", "backend": "transformersjs", "reasoningEffort": "minimal", "temperature": 0.1, "toolPolicy": "none", "outputMode": "text" }, "metadata": { "openRouter": "cpu-local", "funcxBackend": "transformersjs", "envModelId": "TRANSFORMERS_JS_MODEL_ID", "envCacheDir": "TRANSFORMERS_JS_CACHE_DIR", "defaultModelId": "Xenova/distilbart-cnn-6-6", "documentedExampleModelId": "Xenova/distilbart-cnn-6-6", "optionalDep": "@huggingface/transformers", "device": "cpu", "billingModel": "self_hosted" }, "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:local", "cpu:self-hosted" ] }, "modelStatus": "verified", "deploymentType": "selfHosted" } }, "fallbackChoiceOrder": [ "llama_cpp_gguf", "transformersjs_default" ], "requiredCapabilities": { "reasoning": "minimal", "cost": "low", "inputModalities": [ "text" ], "outputModalities": [ "text" ] }, "catalogLane": "text" }, "coding": { "profile": "coding", "displayName": "Coding", "description": "Model selection optimized for software engineering: implementation, refactors, reviews, and debugging.", "category": "engineering", "tags": [ "cod", "eng", "agt", "wf", "eco", "ds", "mmx" ], "aliases": [ "dev", "developer", "software", "programming", "code" ], "defaultChoice": "default", "requiredCapabilities": { "reasoning": "high", "coding": true, "context": "large", "quality": "high", "inputModalities": [ "text" ], "outputModalities": [ "text" ] }, "choices": { "default": { "choice": "default", "displayName": "Coding Default (DeepSeek V4 Flash)", "description": "First choice for agentic coding loops; escalate to deepseek_pro on failure.", "provider": "deepseek", "modelId": "deepseek-v4-flash", "modelStatus": "verified", "runtime": { "instructionTier": "default", "backend": "openrouter", "reasoningEffort": "high", "toolPolicy": "read-only", "outputMode": "text" }, "reason": "Best cost/performance for most coding, patches, and multi-step agent steps.", "verification": { "lastVerifiedAt": "2026-06-01T12:00:00Z", "sourceRefs": [ "vendor:deepseek", "openrouter:catalog" ] } }, "deepseek_pro": { "choice": "deepseek_pro", "displayName": "DeepSeek V4 Pro (escalation)", "description": "Escalation tier when Flash fails or for hard architecture and bug investigation.", "provider": "deepseek", "modelId": "deepseek-v4-pro", "modelStatus": "verified", "runtime": { "instructionTier": "reasoning", "backend": "openrouter", "reasoningEffort": "high", "toolPolicy": "read-only", "outputMode": "text" }, "reason": "Use after Flash struggles on complex refactors, debugging, or long-horizon tasks.", "verification": { "lastVerifiedAt": "2026-06-01T12:00:00Z", "sourceRefs": [ "vendor:deepseek", "openrouter:catalog" ] } }, "claude_coding": { "choice": "claude_coding", "displayName": "Claude Coding", "description": "Premium Claude option for highest instruction-following quality.", "provider": "anthropic", "modelId": "claude-sonnet-4-5", "modelStatus": "verified", "runtime": { "instructionTier": "default", "backend": "openrouter", "reasoningEffort": "high", "toolPolicy": "read-only", "outputMode": "text" }, "reason": "Strong instruction following and code quality when cost is secondary.", "verification": { "lastVerifiedAt": "2026-05-29T20:57:08Z", "sourceRefs": [ "vendor:anthropic", "openrouter:catalog" ] } }, "openai_coding": { "choice": "openai_coding", "displayName": "OpenAI Coding", "description": "OpenAI alternative for coding workloads.", "provider": "openai", "modelId": "gpt-4.1", "modelStatus": "verified", "runtime": { "reasoningEffort": "high", "toolPolicy": "read-only", "outputMode": "text" }, "reason": "Solid alternative for multi-file edits and API-heavy codebases.", "verification": { "lastVerifiedAt": "2026-05-29T20:57:08Z", "sourceRefs": [ "vendor:openai", "openrouter:catalog" ] } }, "minimax_coding": { "choice": "minimax_coding", "displayName": "MiniMax M2.5 Coding", "description": "Cost-efficient coding model with strong SWE-bench performance.", "provider": "minimax", "modelId": "minimax-m2.5", "modelStatus": "verified", "runtime": { "instructionTier": "default", "backend": "openrouter", "reasoningEffort": "high", "toolPolicy": "read-only", "outputMode": "text" }, "reason": "Strong coding benchmarks at a fraction of flagship Claude or GPT pricing.", "verification": { "lastVerifiedAt": "2026-05-30T00:00:00Z", "sourceRefs": [ "vendor:minimax", "openrouter:catalog" ] } } }, "fallbackChoiceOrder": [ "default", "deepseek_pro", "claude_coding", "openai_coding", "minimax_coding" ], "catalogLane": "text" }, "extraction": { "profile": "extraction", "displayName": "Extraction", "description": "Model selection for NER, classification, taxonomy mapping, entity extraction, and document labeling.", "category": "extraction", "tags": [ "xtr", "cls", "str", "json", "bat", "eco" ], "aliases": [ "ner", "classify", "labeling", "entity-extraction", "parse" ], "defaultChoice": "default", "requiredCapabilities": { "structuredOutput": true, "reasoning": "minimal", "cost": "low", "supportsJsonSchema": true, "inputModalities": [ "text" ], "outputModalities": [ "text" ] }, "runtime": { "outputMode": "json", "temperature": 0 }, "choices": { "default": { "choice": "default", "displayName": "Extraction Default", "description": "Fast, low-cost extraction for high-volume labeling pipelines.", "provider": "google", "modelId": "gemini-2.5-flash-lite", "modelStatus": "verified", "runtime": { "reasoningEffort": "minimal", "toolPolicy": "none", "outputMode": "json", "temperature": 0 }, "reason": "Optimized for bulk extraction where latency and cost dominate.", "verification": { "lastVerifiedAt": "2026-05-29T20:57:08Z", "sourceRefs": [ "vendor:google", "openrouter:catalog" ] } }, "anthropic_extraction": { "choice": "anthropic_extraction", "displayName": "Quality Extraction", "description": "Higher-quality extraction when ambiguity or long documents matter.", "provider": "anthropic", "modelId": "claude-haiku-4-5", "modelStatus": "verified", "runtime": { "reasoningEffort": "low", "toolPolicy": "none", "outputMode": "json", "temperature": 0 }, "reason": "Better entity boundaries and taxonomy consistency on messy inputs.", "verification": { "lastVerifiedAt": "2026-05-29T20:57:08Z", "sourceRefs": [ "vendor:anthropic", "openrouter:catalog" ] } } }, "fallbackChoiceOrder": [ "default", "anthropic_extraction" ], "catalogLane": "text" }, "routing": { "profile": "routing", "displayName": "Routing", "description": "Ultra-low-cost models for intent classification, record routing, and workflow branching.", "category": "routing", "tags": [ "rte", "tri", "cls", "str", "eco" ], "aliases": [ "router", "classifier", "intent", "branch", "triage" ], "defaultChoice": "default", "requiredCapabilities": { "reasoning": "minimal", "cost": "low", "latency": "low", "inputModalities": [ "text" ], "outputModalities": [ "text" ] }, "choices": { "default": { "choice": "default", "displayName": "Routing Default", "description": "Cheapest reliable option for high-volume routing decisions.", "provider": "google", "modelId": "gemini-2.5-flash-lite", "modelStatus": "verified", "runtime": { "reasoningEffort": "minimal", "toolPolicy": "none", "outputMode": "json", "temperature": 0 }, "reason": "Minimize spend on per-request routing and orchestration gates.", "verification": { "lastVerifiedAt": "2026-05-29T20:57:08Z", "sourceRefs": [ "vendor:google", "openrouter:catalog" ] } }, "safer_routing": { "choice": "safer_routing", "displayName": "Safer Routing", "description": "Slightly higher cost when routing mistakes are expensive.", "provider": "anthropic", "modelId": "claude-haiku-4-5", "modelStatus": "verified", "runtime": { "reasoningEffort": "low", "toolPolicy": "none", "outputMode": "json", "temperature": 0 }, "reason": "Better instruction adherence for multi-label or nested routing trees.", "verification": { "lastVerifiedAt": "2026-05-29T20:57:08Z", "sourceRefs": [ "vendor:anthropic", "openrouter:catalog" ] } } }, "fallbackChoiceOrder": [ "default", "safer_routing" ], "catalogLane": "text" }, "vol": { "profile": "vol", "displayName": "Volume Agent", "description": "High-volume economic agent lane: Flash for default agent, coding, extraction, and workflows; Pro for hard reasoning and final-quality answers.", "category": "agent", "tags": [ "ds", "eco", "agt", "wf", "cod", "anl", "rsn", "str", "json", "xtr", "cls", "pro", "rev" ], "aliases": [ "agent-volume", "flash-tier", "volume-agent", "final-answer" ], "defaultChoice": "default", "choices": { "default": { "choice": "default", "displayName": "DeepSeek Default (V4 Flash)", "description": "Default app, agent, and workflow model with thinking enabled.", "provider": "deepseek", "modelId": "deepseek-v4-flash", "modelStatus": "verified", "runtime": { "instructionTier": "default", "backend": "openrouter", "reasoningEffort": "medium", "toolPolicy": "write-with-approval", "outputMode": "text" }, "reason": "Cheap, fast, and strong enough for most coding, extraction, routing, summaries, and agent steps (1M context).", "verification": { "lastVerifiedAt": "2026-06-01T12:00:00Z", "sourceRefs": [ "vendor:deepseek", "openrouter:catalog", "vendor:deepseek-api-docs" ] }, "secondary": { "kind": "choice", "choice": "pro", "reason": "Escalation to Pro when Flash default is insufficient for agent steps." } }, "flash": { "choice": "flash", "displayName": "V4 Flash (non-thinking)", "description": "Classification, extraction, and JSON tasks without expensive reasoning.", "provider": "deepseek", "modelId": "deepseek-v4-flash", "modelStatus": "verified", "runtime": { "instructionTier": "default", "backend": "openrouter", "reasoningEffort": "minimal", "toolPolicy": "none", "outputMode": "schema", "temperature": 0 }, "reason": "Structured extraction and high-volume tasks that do not need deep reasoning.", "verification": { "lastVerifiedAt": "2026-06-01T12:00:00Z", "sourceRefs": [ "vendor:deepseek", "openrouter:catalog" ] } }, "pro": { "choice": "pro", "displayName": "V4 Pro (hard reasoning)", "description": "Complex reasoning, architecture, bug investigation, and long-context analysis.", "provider": "deepseek", "modelId": "deepseek-v4-pro", "modelStatus": "verified", "runtime": { "instructionTier": "reasoning", "backend": "openrouter", "reasoningEffort": "high", "toolPolicy": "read-only", "outputMode": "text" }, "reason": "Escalation tier for agentic coding loops and hard multi-step reasoning at a fraction of Opus or GPT-5.5 cost.", "verification": { "lastVerifiedAt": "2026-06-01T12:00:00Z", "sourceRefs": [ "vendor:deepseek", "openrouter:catalog", "vendor:deepseek-api-docs" ] } }, "final": { "choice": "final", "displayName": "V4 Pro (final answer)", "description": "Professional-grade final review and answer nodes.", "provider": "deepseek", "modelId": "deepseek-v4-pro", "modelStatus": "verified", "runtime": { "instructionTier": "reasoning", "backend": "openrouter", "reasoningEffort": "high", "toolPolicy": "none", "outputMode": "text" }, "reason": "Best place to spend DeepSeek quality budget on final outputs.", "verification": { "lastVerifiedAt": "2026-06-01T12:00:00Z", "sourceRefs": [ "vendor:deepseek", "openrouter:catalog" ] } } }, "fallbackChoiceOrder": [ "default", "flash", "pro", "final" ], "requiredCapabilities": { "reasoning": "medium", "coding": true, "context": "xlarge", "cost": "low", "supportsTools": true, "inputModalities": [ "text" ], "outputModalities": [ "text" ] }, "catalogLane": "text" }, "sum": { "profile": "sum", "displayName": "Summarization", "description": "Optimized for information compression and document triage.", "category": "analysis", "tags": [ "sum", "anl", "tri", "wf" ], "aliases": [ "triage", "digest", "tl-dr" ], "defaultChoice": "default", "choices": { "default": { "choice": "default", "displayName": "Fast Digest", "description": "Quickly summarizes standard documents and meeting transcripts.", "provider": "anthropic", "modelId": "claude-haiku-4-5", "runtime": { "instructionTier": "default", "backend": "openrouter", "reasoningEffort": "low", "toolPolicy": "none", "outputMode": "text" }, "reason": "Fast and capable at reading comprehension and triage summaries.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:anthropic", "openrouter:catalog" ] }, "modelStatus": "verified" }, "mega_context": { "choice": "mega_context", "displayName": "Mega Context Summarization", "description": "Compressing entire repositories, books, or multi-year document histories.", "provider": "google", "modelId": "gemini-3.5-flash", "runtime": { "reasoningEffort": "high", "toolPolicy": "none", "outputMode": "text" }, "reason": "When input exceeds ~200k tokens, Gemini 3.5 Flash long context handles needle-in-haystack retrieval best.", "verification": { "lastVerifiedAt": "2026-05-29T17:08:03Z", "sourceRefs": [ "vendor:google", "openrouter:catalog" ] }, "modelStatus": "verified" } }, "fallbackChoiceOrder": [ "default", "mega_context" ], "requiredCapabilities": { "reasoning": "low", "cost": "low", "context": "large", "inputModalities": [ "text" ], "outputModalities": [ "text" ] }, "catalogLane": "text" } } }