# AI FinOps Cost Oracle - Full Technical & Mathematical Reference Canonical Source: https://Thorlan04.github.io/ai-finops-cost-oracle/llms-full.txt Published by: AI FinOps Cost Oracle Engineering Collective License: Open Data / Permissive Free Reference ## 1. Deterministic Cost Formulas ### 1.1 Large Language Model Invoicing Formula For any model m with input price per token P_in and output price per token P_out: MonthlyCost(m) = TotalMonthlyCalls * [ (AvgInputTokens * P_in) + (AvgOutputTokens * P_out) ] Where: TotalMonthlyCalls = DailyRequests * ActiveUsers * 30.4375 ### 1.2 Automation Break-Even Arbitrage Formula Let C_vps be the fixed monthly VPS cost (Hostinger KVM2 = $7.50 / 6.99 €). Let T_zapier be the effective marginal cost per Zapier task ($0.038). BreakEvenVolume = C_vps / T_zapier = 7.50 / 0.038 = 197.37 tasks per month. Any workflow running more than 200 tasks per month achieves a positive ROI when migrated to self-hosted n8n. ## 2. Complete Pricing Matrix JSON Dataset ```json { "schema_version": "1.0.0", "generated_at": "2026-11-16T15:25:35.773316+00:00", "source": "AI FinOps Cost Oracle Automated Multi-Cloud Engine", "base_currency": "USD", "eur_usd_rate": 1.08, "models": { "anthropic/claude-3-7-sonnet": { "provider": "Anthropic", "model_name": "Claude 3.7 Sonnet (Hybrid Reasoning)", "input_cost_per_token": 3e-06, "output_cost_per_token": 1.5e-05, "input_cost_per_million": 3.0, "output_cost_per_million": 15.0, "context_window": 200000, "max_output_tokens": 64000, "category": "frontier_hybrid", "description": "Anthropic frontier hybrid reasoning model combining instant execution with controllable extended thinking tokens.", "last_synced_from": "LiteLLM-Live" }, "openai/o3-mini": { "provider": "OpenAI", "model_name": "OpenAI o3-mini", "input_cost_per_token": 1.1e-06, "output_cost_per_token": 4.4e-06, "input_cost_per_million": 1.1, "output_cost_per_million": 4.4, "context_window": 200000, "max_output_tokens": 100000, "category": "reasoning_ultra", "description": "Next-gen high-efficiency STEM and coding reasoning model replacing o1-mini at lower cost and higher speed.", "last_synced_from": "LiteLLM-Live" }, "openai/o3": { "provider": "OpenAI", "model_name": "OpenAI o3", "input_cost_per_token": 2e-06, "output_cost_per_token": 8e-06, "input_cost_per_million": 2.0, "output_cost_per_million": 8.0, "context_window": 200000, "max_output_tokens": 100000, "category": "reasoning_frontier", "description": "State-of-the-art general reasoning model superseding o1-preview with superior agentic performance.", "last_synced_from": "LiteLLM-Live" }, "openai/gpt-4-5": { "provider": "OpenAI", "model_name": "GPT-4.5 (Super-Scale Knowledge)", "input_cost_per_token": 7.5e-05, "output_cost_per_token": 0.00015, "input_cost_per_million": 75.0, "output_cost_per_million": 150.0, "context_window": 128000, "max_output_tokens": 16384, "category": "massive_knowledge", "description": "Colossal world knowledge and nuanced qualitative synthesis model for mission-critical enterprise decisions." }, "google/gemini-2-0-flash": { "provider": "Google", "model_name": "Gemini 2.0 Flash", "input_cost_per_token": 1e-07, "output_cost_per_token": 4e-07, "input_cost_per_million": 0.1, "output_cost_per_million": 0.4, "context_window": 1048576, "max_output_tokens": 8192, "category": "cost_optimized", "description": "Second-generation multimodal foundation model with native tool execution and sub-second latency." }, "google/gemini-2-5-pro": { "provider": "Google", "model_name": "Gemini 2.5 Pro", "input_cost_per_token": 1.25e-06, "output_cost_per_token": 1e-05, "input_cost_per_million": 1.25, "output_cost_per_million": 10.0, "context_window": 1048576, "max_output_tokens": 16384, "category": "flagship", "description": "Advanced multi-modal reasoning with 2M token context, agentic planning, and codebase ingestion.", "last_synced_from": "LiteLLM-Live" }, "deepseek/deepseek-reasoner": { "provider": "DeepSeek", "model_name": "DeepSeek R1 (Reasoner)", "input_cost_per_token": 2.8e-07, "output_cost_per_token": 4.2e-07, "input_cost_per_million": 0.28, "output_cost_per_million": 0.42, "context_window": 131072, "max_output_tokens": 8192, "category": "reasoning_open", "description": "Groundbreaking open reasoning model matching proprietary frontier reasoning at 90-96% lower cost.", "last_synced_from": "LiteLLM-Live" }, "deepseek/deepseek-chat": { "provider": "DeepSeek", "model_name": "DeepSeek V3 (Chat)", "input_cost_per_token": 2.8e-07, "output_cost_per_token": 4.2e-07, "input_cost_per_million": 0.28, "output_cost_per_million": 0.42, "context_window": 131072, "max_output_tokens": 8192, "category": "hyper_efficient", "description": "671B MoE architecture delivering frontier coding and general reasoning at ultra-low price.", "last_synced_from": "LiteLLM-Live" }, "openai/gpt-4o": { "provider": "OpenAI", "model_name": "GPT-4o (Omni)", "input_cost_per_token": 2.5e-06, "output_cost_per_token": 1e-05, "input_cost_per_million": 2.5, "output_cost_per_million": 10.0, "context_window": 128000, "max_output_tokens": 16384, "category": "flagship", "description": "OpenAI general multimodal workhorse model.", "last_synced_from": "LiteLLM-Live" }, "openai/gpt-4o-mini": { "provider": "OpenAI", "model_name": "GPT-4o Mini", "input_cost_per_token": 1.5e-07, "output_cost_per_token": 6e-07, "input_cost_per_million": 0.15, "output_cost_per_million": 0.6, "context_window": 128000, "max_output_tokens": 16384, "category": "cost_optimized", "description": "Fast, economical multimodal model for high-frequency routing.", "last_synced_from": "LiteLLM-Live" }, "anthropic/claude-3-5-sonnet": { "provider": "Anthropic", "model_name": "Claude 3.5 Sonnet", "input_cost_per_token": 3e-06, "output_cost_per_token": 1.5e-05, "input_cost_per_million": 3.0, "output_cost_per_million": 15.0, "context_window": 200000, "max_output_tokens": 8192, "category": "flagship", "description": "High capability programming and reasoning predecessor to Claude 3.7." }, "anthropic/claude-3-5-haiku": { "provider": "Anthropic", "model_name": "Claude 3.5 Haiku", "input_cost_per_token": 1e-06, "output_cost_per_token": 5e-06, "input_cost_per_million": 1.0, "output_cost_per_million": 5.0, "context_window": 200000, "max_output_tokens": 8192, "category": "cost_optimized", "description": "Sub-second lightweight execution from Anthropic." }, "mistral/mistral-large": { "provider": "Mistral AI", "model_name": "Mistral Large 2", "input_cost_per_token": 2e-06, "output_cost_per_token": 6e-06, "input_cost_per_million": 2.0, "output_cost_per_million": 6.0, "context_window": 128000, "max_output_tokens": 8192, "category": "flagship", "description": "European sovereign enterprise frontier model." }, "mistral/codestral": { "provider": "Mistral AI", "model_name": "Codestral 25.01", "input_cost_per_token": 2e-07, "output_cost_per_token": 6e-07, "input_cost_per_million": 0.2, "output_cost_per_million": 0.6, "context_window": 256000, "max_output_tokens": 8192, "category": "code_specialist", "description": "Specialized generative coding and fill-in-the-middle completion model with 256k context." }, "meta/llama-3.3-70b-instruct": { "provider": "Meta (via Groq / Together / DeepInfra)", "model_name": "Llama 3.3 70B Instruct", "input_cost_per_token": 5.9e-07, "output_cost_per_token": 7.9e-07, "input_cost_per_million": 0.59, "output_cost_per_million": 0.79, "context_window": 128000, "max_output_tokens": 4096, "category": "open_source_cloud", "description": "Open weights standard flagship model with near-405B benchmark parity." } }, "automation_platforms": { "zapier": { "name": "Zapier (Cloud SaaS)", "billing_model": "per_task_tiered", "free_tier_ops": 100, "plans": [ { "name": "Free", "monthly_price_usd": 0, "included_ops": 100, "overage_per_1000": 0 }, { "name": "Starter", "monthly_price_usd": 29.99, "included_ops": 750, "overage_per_1000": 45.0 }, { "name": "Professional", "monthly_price_usd": 73.5, "included_ops": 2000, "overage_per_1000": 38.0 }, { "name": "Team", "monthly_price_usd": 103.5, "included_ops": 2000, "overage_per_1000": 34.0 }, { "name": "High Volume 50k", "monthly_price_usd": 448.5, "included_ops": 50000, "overage_per_1000": 15.0 }, { "name": "Enterprise 100k", "monthly_price_usd": 898.0, "included_ops": 100000, "overage_per_1000": 12.0 } ], "average_cost_per_task_usd": 0.038, "notes": "Premium proprietary pricing. Steep penalties on multi-step workflows where each step counts as 1 task." }, "make": { "name": "Make (formerly Integromat)", "billing_model": "per_operation_tiered", "free_tier_ops": 1000, "plans": [ { "name": "Free", "monthly_price_usd": 0, "included_ops": 1000, "overage_per_1000": 0 }, { "name": "Core", "monthly_price_usd": 10.59, "included_ops": 10000, "overage_per_1000": 1.2 }, { "name": "Pro", "monthly_price_usd": 18.82, "included_ops": 10000, "overage_per_1000": 1.8 }, { "name": "Teams", "monthly_price_usd": 34.12, "included_ops": 10000, "overage_per_1000": 3.4 }, { "name": "Pro 50k", "monthly_price_usd": 68.0, "included_ops": 50000, "overage_per_1000": 1.36 }, { "name": "Pro 100k", "monthly_price_usd": 129.0, "included_ops": 100000, "overage_per_1000": 1.29 } ], "average_cost_per_task_usd": 0.0015, "notes": "Cost-effective visual workflow editor. 60% to 80% cheaper than Zapier for high-volume routing." }, "n8n_cloud": { "name": "n8n (Managed Cloud)", "billing_model": "per_workflow_execution", "free_tier_ops": 0, "plans": [ { "name": "Starter", "monthly_price_usd": 22.0, "included_executions": 2500, "overage_per_1000": 8.0 }, { "name": "Pro", "monthly_price_usd": 55.0, "included_executions": 10000, "overage_per_1000": 5.0 }, { "name": "Enterprise", "monthly_price_usd": 180.0, "included_executions": 50000, "overage_per_1000": 3.0 } ], "notes": "Billed per entire workflow execution rather than individual step operations. Massive efficiency for complex trees." }, "n8n_self_hosted_hostinger": { "name": "n8n Self-Hosted (Hostinger KVM2 VPS)", "billing_model": "fixed_infrastructure", "monthly_price_usd": 7.5, "monthly_price_eur": 6.99, "included_ops": 10000000, "vps_specs": "2 vCPU, 8 GB RAM, 100 GB NVMe, 8 TB Bandwidth, Docker Containerized", "notes": "100% Unlimited task execution. Total cost remains constant regardless of 10k or 1,000k tasks/month." }, "n8n_self_hosted_digitalocean": { "name": "n8n Self-Hosted (DigitalOcean Basic Droplet)", "billing_model": "fixed_infrastructure", "monthly_price_usd": 12.0, "monthly_price_eur": 11.1, "included_ops": 10000000, "vps_specs": "2 vCPU, 2 GB RAM, 60 GB SSD, 3 TB Bandwidth, 1-Click Docker Marketplace", "notes": "Enterprise SLA uptime, 1-click Docker marketplace deploy, unlimited workflow executions." } }, "vector_databases": { "supabase": { "name": "Supabase pgvector (PostgreSQL)", "free_tier_vectors": 50000, "pro_plan_monthly_usd": 25.0, "included_storage_gb": 8, "included_bandwidth_gb": 250, "overage_storage_per_gb_usd": 0.125, "read_query_cost_per_million_usd": 0.0, "notes": "Relational data + pgvector in one place. Predictable flat compute tiers without per-read microcharges." }, "pinecone": { "name": "Pinecone Serverless", "free_tier_vectors": 100000, "base_fee_monthly_usd": 0.0, "read_unit_cost_per_million_usd": 0.25, "write_unit_cost_per_million_usd": 0.75, "storage_per_gb_monthly_usd": 0.33, "notes": "Pure serverless vector index. Pay-as-you-go per Read/Write unit, ideal for bursty sparse query traffic." }, "qdrant_cloud": { "name": "Qdrant Cloud Managed", "free_tier_vectors": 100000, "cluster_base_monthly_usd": 25.0, "notes": "High performance Rust-based vector engine with payload filtering and hybrid search." } } } ```