mcp-ai-agent-guidelines/dist/tools/config/models.yaml

A comprehensive Model Context Protocol server providing advanced tools, resources, and prompts for implementing AI agent best practices

# AI Model Definitions
# This file contains the definitions of supported AI models for the MCP AI Agent Guidelines
# Source: https://docs.github.com/en/copilot/reference/ai-models/supported-models
# Last updated: 2025-12-15
#
# Schema Fields:
# Required Fields:
#   - name: Display name of the model (string)
#   - provider: Provider/company name (string)
#   - pricingTier: One of "premium", "mid-tier", or "budget"
#   - contextTokens: Context window size in tokens (integer)
#   - baseScore: Base scoring value 0-100 (integer)
#   - capabilities: Array of capability strings
#   - strengths: Array of strength descriptions
#   - limitations: Array of limitation descriptions
#   - specialFeatures: Array of special feature descriptions
#   - pricing: Human-readable pricing description
#
# Optional Fields (NEW):
#   - modes: Object with boolean flags for supported modes
#     - agent: Supports agent mode (boolean)
#     - reasoning: Deep reasoning capability (boolean)
#     - vision: Multimodal image input (boolean)
#     - chat: Chat completions (boolean)
#     - edit: Code editing mode (boolean)
#     - completions: Inline suggestions (boolean)
#   - taskArea: Task categorization - "general-purpose", "deep-reasoning", "fast-simple", or "visual"
#   - multiplier: Premium request multiplier (number, e.g., 1.0, 1.5, 2.0)
#   - status: Model status - "ga", "preview", "beta", or "retired"
#   - documentationUrl: Link to official model documentation (string)

# Default model configuration
# This model is used as the default across all prompt builders
# Change this value to easily switch the default model project-wide
defaultModel: "GPT-5-Codex"

models:
  - name: "GPT-4.1"
    provider: "OpenAI"
    pricingTier: "mid-tier"
    contextTokens: 128000
    baseScore: 52
    capabilities:
      - "reasoning"
      - "code"
      - "speed"
      - "multimodal"
    strengths:
      - "Fast, accurate code completions"
      - "General-purpose coding"
      - "Writing tasks"
      - "Wide ecosystem"
    limitations:
      - "Context window limitations"
    specialFeatures:
      - "Agent mode"
      - "Vision capabilities"
      - "Function calling"
    pricing: "Mid-tier ($5-10/1M tokens)"
    # New metadata fields
    modes:
      agent: true
      reasoning: false
      vision: true
      chat: true
      edit: true
      completions: true
    taskArea: "general-purpose"
    multiplier: 1.0
    status: "ga"
    documentationUrl: "https://openai.com/index/gpt-4-1/"

  - name: "GPT-5"
    provider: "OpenAI"
    pricingTier: "premium"
    contextTokens: 128000
    baseScore: 54
    capabilities:
      - "reasoning"
      - "code"
      - "multimodal"
    strengths:
      - "Multi-step problem solving"
      - "Architecture-level code analysis"
      - "Deep reasoning"
    limitations:
      - "Higher cost"
      - "Context limitations"
    specialFeatures:
      - "Advanced reasoning"
      - "Complex debugging"
    pricing: "Premium ($10-20/1M tokens)"
    # New metadata fields
    modes:
      agent: true
      reasoning: true
      vision: true
      chat: true
      edit: true
      completions: false
    taskArea: "deep-reasoning"
    multiplier: 1.5
    status: "ga"
    documentationUrl: "https://openai.com/gpt-5/"

  - name: "GPT-5.1"
    provider: "OpenAI"
    pricingTier: "premium"
    contextTokens: 128000
    baseScore: 56
    capabilities:
      - "reasoning"
      - "code"
    strengths:
      - "Multi-step problem solving"
      - "Architecture-level code analysis"
      - "Deep reasoning"
      - "Superior debugging capabilities"
    limitations:
      - "Higher cost"
      - "Slower inference"
    specialFeatures:
      - "Advanced reasoning chains"
      - "Complex problem solving"
      - "Architecture-level analysis"
    pricing: "Premium ($15-30/1M tokens)"
    modes:
      agent: true
      reasoning: true
      vision: false
      chat: true
      edit: true
      completions: false
    taskArea: "deep-reasoning"
    multiplier: 2.0
    status: "ga"
    documentationUrl: "https://openai.com/gpt-5/"

  - name: "GPT-5.2"
    provider: "OpenAI"
    pricingTier: "premium"
    contextTokens: 128000
    baseScore: 57
    capabilities:
      - "reasoning"
      - "code"
      - "multimodal"
    strengths:
      - "Advanced multi-step problem solving"
      - "Superior architecture-level code analysis"
      - "Enhanced deep reasoning"
      - "Complex debugging workflows"
    limitations:
      - "Higher cost"
      - "Slower inference"
      - "Context limitations"
    specialFeatures:
      - "Advanced reasoning chains"
      - "Complex problem solving"
      - "Architecture-level analysis"
      - "Agent mode"
    pricing: "Premium ($20-40/1M tokens)"
    modes:
      agent: true
      reasoning: true
      vision: false
      chat: true
      edit: true
      completions: false
    taskArea: "deep-reasoning"
    multiplier: 2.5
    status: "ga"
    documentationUrl: "https://openai.com/gpt-5/"

  - name: "GPT-5-Codex"
    provider: "OpenAI"
    pricingTier: "mid-tier"
    contextTokens: 128000
    baseScore: 54
    capabilities:
      - "code"
      - "reasoning"
      - "speed"
    strengths:
      - "Higher-quality code on complex tasks"
      - "Features, tests, debugging, refactors"
      - "Works without lengthy instructions"
    limitations:
      - "Moderate cost"
    specialFeatures:
      - "Agent mode"
      - "Optimized for coding"
    pricing: "Mid-tier ($5-10/1M tokens)"
    modes:
      agent: true
      reasoning: false
      vision: false
      chat: true
      edit: true
      completions: true
    taskArea: "general-purpose"
    multiplier: 1.0
    status: "ga"
    documentationUrl: "https://cdn.openai.com/pdf/97cc5669-7a25-4e63-b15f-5fd5bdc4d149/gpt-5-codex-system-card.pdf"

  - name: "GPT-5.1-Codex"
    provider: "OpenAI"
    pricingTier: "mid-tier"
    contextTokens: 128000
    baseScore: 55
    capabilities:
      - "code"
      - "reasoning"
      - "speed"
    strengths:
      - "Optimized for complex coding tasks"
      - "Enhanced reasoning for code"
      - "Features, tests, debugging, refactors"
      - "Works without lengthy instructions"
    limitations:
      - "Moderate cost"
    specialFeatures:
      - "Agent mode"
      - "Enhanced Codex optimization"
      - "Improved code generation"
    pricing: "Mid-tier ($8-15/1M tokens)"
    modes:
      agent: true
      reasoning: true
      vision: false
      chat: true
      edit: true
      completions: true
    taskArea: "general-purpose"
    multiplier: 1.0
    status: "ga"
    documentationUrl: "https://cdn.openai.com/pdf/97cc5669-7a25-4e63-b15f-5fd5bdc4d149/gpt-5-codex-system-card.pdf"

  - name: "GPT-5.1-Codex-Mini"
    provider: "OpenAI"
    pricingTier: "mid-tier"
    contextTokens: 128000
    baseScore: 54
    capabilities:
      - "code"
      - "reasoning"
      - "speed"
    strengths:
      - "Fast, accurate code completions"
      - "Reasoning-enhanced code generation"
      - "Lightweight and responsive"
      - "Good for iterative development"
    limitations:
      - "Moderate cost"
    specialFeatures:
      - "Agent mode"
      - "Reasoning capability"
      - "Codex optimization"
    pricing: "Mid-tier ($5-10/1M tokens)"
    modes:
      agent: true
      reasoning: true
      vision: false
      chat: true
      edit: true
      completions: true
    taskArea: "general-purpose"
    multiplier: 1.0
    status: "ga"
    documentationUrl: "https://cdn.openai.com/pdf/97cc5669-7a25-4e63-b15f-5fd5bdc4d149/gpt-5-codex-system-card.pdf"

  - name: "GPT-5 mini"
    provider: "OpenAI"
    pricingTier: "mid-tier"
    contextTokens: 128000
    baseScore: 53
    capabilities:
      - "reasoning"
      - "speed"
      - "multimodal"
      - "code"
    strengths:
      - "Fast, accurate code completions"
      - "Deep reasoning with faster responses"
      - "Ideal for interactive sessions"
      - "Step-by-step code analysis"
    limitations:
      - "Moderate cost"
    specialFeatures:
      - "Agent mode"
      - "Reasoning capability"
      - "Vision support"
    pricing: "Mid-tier ($3-6/1M tokens)"
    modes:
      agent: true
      reasoning: true
      vision: true
      chat: true
      edit: true
      completions: true
    taskArea: "general-purpose"
    multiplier: 1.0
    status: "ga"
    documentationUrl: "https://cdn.openai.com/gpt-5-system-card.pdf"

  - name: "Claude Opus 4.1"
    provider: "Anthropic"
    pricingTier: "premium"
    contextTokens: 200000
    baseScore: 55
    capabilities:
      - "reasoning"
      - "large-context"
      - "multimodal"
      - "code"
    strengths:
      - "Complex problem-solving"
      - "Sophisticated reasoning"
      - "Vision capabilities"
    limitations:
      - "Higher cost"
      - "Slower inference"
    specialFeatures:
      - "Constitutional AI"
      - "Vision support"
      - "Deep reasoning"
    pricing: "Premium ($15-30/1M tokens)"
    # New metadata fields
    modes:
      agent: true
      reasoning: true
      vision: true
      chat: true
      edit: true
      completions: false
    taskArea: "deep-reasoning"
    multiplier: 2.0
    status: "ga"
    documentationUrl: "https://www.anthropic.com/claude/opus"

  - name: "Claude Haiku 4.5"
    provider: "Anthropic"
    pricingTier: "budget"
    contextTokens: 200000
    baseScore: 49
    capabilities:
      - "speed"
      - "cost"
      - "code"
    strengths:
      - "Fast, reliable answers"
      - "Lightweight coding questions"
      - "Small tasks and utility code"
      - "Quick syntax help"
    limitations:
      - "Limited reasoning depth"
    specialFeatures:
      - "Agent mode"
      - "Ultra-fast responses"
      - "Cost efficient"
    pricing: "Budget ($0.25-1.25/1M tokens)"
    modes:
      agent: true
      reasoning: false
      vision: false
      chat: true
      edit: true
      completions: true
    taskArea: "fast-simple"
    multiplier: 0.5
    status: "ga"
    documentationUrl: "https://www.anthropic.com/claude/haiku"

  - name: "Claude Sonnet 4.5"
    provider: "Anthropic"
    pricingTier: "mid-tier"
    contextTokens: 200000
    baseScore: 54
    capabilities:
      - "reasoning"
      - "code"
      - "large-context"
    strengths:
      - "Complex problem-solving challenges"
      - "Sophisticated reasoning"
      - "Perfectly balanced for coding workflows"
    limitations:
      - "Moderate cost"
    specialFeatures:
      - "Agent mode"
      - "Superior reasoning"
      - "Coding optimization"
    pricing: "Mid-tier ($3-6/1M tokens)"
    modes:
      agent: true
      reasoning: true
      vision: false
      chat: true
      edit: true
      completions: true
    taskArea: "general-purpose"
    multiplier: 1.0
    status: "ga"
    documentationUrl: "https://assets.anthropic.com/m/12f214efcc2f457a/original/Claude-Sonnet-4-5-System-Card.pdf"

  - name: "Claude Sonnet 4"
    provider: "Anthropic"
    pricingTier: "mid-tier"
    contextTokens: 200000
    baseScore: 53
    capabilities:
      - "reasoning"
      - "speed"
      - "multimodal"
      - "code"
    strengths:
      - "Performance and practicality"
      - "Balanced for coding workflows"
    limitations:
      - "Moderate cost"
    specialFeatures:
      - "Agent mode"
      - "Vision support"
      - "Balanced speed/quality"
    pricing: "Mid-tier ($3-6/1M tokens)"
    # New metadata fields
    modes:
      agent: true
      reasoning: true
      vision: true
      chat: true
      edit: true
      completions: true
    taskArea: "general-purpose"
    multiplier: 1.0
    status: "ga"
    documentationUrl: "https://www.anthropic.com/claude/sonnet"

  - name: "Gemini 2.5 Pro"
    provider: "Google"
    pricingTier: "mid-tier"
    contextTokens: 2000000
    baseScore: 54
    capabilities:
      - "reasoning"
      - "multimodal"
      - "large-context"
      - "code"
    strengths:
      - "Complex code generation"
      - "Debugging"
      - "Research workflows"
      - "Massive context"
    limitations:
      - "Limited third-party integration"
    specialFeatures:
      - "2M token context"
      - "Vision support"
      - "Google integration"
    pricing: "Competitive ($1-3/1M tokens)"
    # New metadata fields
    modes:
      agent: true
      reasoning: true
      vision: true
      chat: true
      edit: true
      completions: true
    taskArea: "general-purpose"
    multiplier: 1.0
    status: "ga"
    documentationUrl: "https://ai.google.dev/gemini-api/docs"

  - name: "Gemini 2.0 Flash"
    provider: "Google"
    pricingTier: "budget"
    contextTokens: 1000000
    baseScore: 50
    capabilities:
      - "speed"
      - "multimodal"
      - "cost"
    strengths:
      - "Real-time responses"
      - "Visual reasoning"
      - "UI and diagram tasks"
    limitations:
      - "Limited reasoning depth"
    specialFeatures:
      - "Vision capabilities"
      - "Fast inference"
      - "Low latency"
    pricing: "Budget ($0.25-1/1M tokens)"
    modes:
      agent: false
      reasoning: false
      vision: true
      chat: true
      edit: true
      completions: true
    taskArea: "fast-simple"
    multiplier: 0.5
    status: "ga"
    documentationUrl: "https://ai.google.dev/gemini-api/docs"

  - name: "Grok Code Fast 1"
    provider: "xAI"
    pricingTier: "mid-tier"
    contextTokens: 128000
    baseScore: 52
    capabilities:
      - "code"
      - "speed"
      - "reasoning"
    strengths:
      - "Specialized for coding tasks"
      - "Code generation and debugging"
      - "Fast, accurate completions"
      - "Multi-language support"
    limitations:
      - "Moderate cost"
    specialFeatures:
      - "Agent mode"
      - "Optimized for code"
      - "Complimentary access (limited time)"
    pricing: "Mid-tier ($3-6/1M tokens, currently complimentary)"
    modes:
      agent: true
      reasoning: false
      vision: false
      chat: true
      edit: true
      completions: true
    taskArea: "general-purpose"
    multiplier: 0.0
    status: "preview"
    documentationUrl: "https://data.x.ai/2025-08-20-grok-4-model-card.pdf"

  - name: "Qwen2.5"
    provider: "Alibaba"
    pricingTier: "mid-tier"
    contextTokens: 128000
    baseScore: 51
    capabilities:
      - "code"
      - "reasoning"
    strengths:
      - "Code generation"
      - "Code repair and debugging"
      - "Strong reasoning capabilities"
    limitations:
      - "Moderate cost"
      - "Less established ecosystem"
    specialFeatures:
      - "Reasoning mode"
      - "Multi-language code support"
    pricing: "Mid-tier ($3-6/1M tokens)"
    modes:
      agent: false
      reasoning: true
      vision: false
      chat: true
      edit: true
      completions: true
    taskArea: "general-purpose"
    multiplier: 1.0
    status: "ga"
    documentationUrl: "https://arxiv.org/pdf/2409.12186"

  - name: "Raptor mini"
    provider: "Meta"
    pricingTier: "budget"
    contextTokens: 128000
    baseScore: 50
    capabilities:
      - "speed"
      - "code"
      - "cost"
    strengths:
      - "Fast, accurate inline suggestions"
      - "Quick code completions"
      - "Lightweight and responsive"
    limitations:
      - "Limited reasoning depth"
      - "Best for simple tasks"
    specialFeatures:
      - "Agent mode"
      - "Optimized for inline completions"
      - "Ultra-low latency"
    pricing: "Budget ($0.25-1.25/1M tokens)"
    modes:
      agent: true
      reasoning: false
      vision: false
      chat: true
      edit: true
      completions: true
    taskArea: "general-purpose"
    multiplier: 0.5
    status: "preview"
    documentationUrl: "https://ai.meta.com/raptor"

# Requirement keywords for capability matching
requirementKeywords:
  reasoning:
    - "analysis"
    - "reasoning"
    - "complex"
    - "chain-of-thought"
    - "logic"
  code:
    - "code"
    - "programming"
    - "development"
    - "refactor"
    - "api"
  large-context:
    - "large"
    - "document"
    - "context"
    - "long"
    - "200k"
    - "2m"
  speed:
    - "fast"
    - "real-time"
    - "quick"
    - "low-latency"
    - "interactive"
  multimodal:
    - "image"
    - "visual"
    - "multimodal"
    - "audio"
    - "video"
  safety:
    - "safety"
    - "reliable"
    - "production"
    - "alignment"
    - "guardrail"
  cost:
    - "cheap"
    - "cost"
    - "budget"
    - "economical"
    - "low-cost"

# Capability weights for scoring
capabilityWeights:
  reasoning: 18
  code: 14
  large-context: 22
  speed: 16
  multimodal: 18
  safety: 15
  cost: 15

# Budget adjustments for scoring
budgetAdjustments:
  low:
    bonus:
      - "budget"
      - "mid-tier"
    penalty:
      - "premium"
  medium:
    bonus:
      - "mid-tier"
    penalty: []
  high:
    bonus:
      - "premium"
    penalty: []

# Scoring constants
budgetBonus: 12
budgetPenalty: 14