Add Model Price

Purpose

Guide for adding new LLM model pricing entries to Langfuse's default model prices configuration. This enables accurate cost tracking across different model providers and deployment platforms.

When to Use This Skill

Automatically activates when:

• Adding a new model to the pricing database
• Updating model pricing information
• Working with default-model-prices.json
• Creating model matchPattern regex
• Configuring pricingTiers or tiered pricing
• Adding prices for Claude, OpenAI, Anthropic, Google, Gemini, AWS Bedrock, Azure, or Vertex AI models

Quick Start Checklist

• Gather model info: Fetch official pricing from provider documentation URL
• Generate UUID: Run uuidgen for the model entry ID (use lowercase)
• Create matchPattern: Regex covering all provider formats
• Define pricingTiers: At minimum, one default tier with standard prices
• Add pricing entry: Insert into /worker/src/constants/default-model-prices.json
• Add to LLM types: Add model to /packages/shared/src/server/llm/types.ts (for playground/LLM-as-judge)
• Validate JSON: Run jq . default-model-prices.json to verify syntax

File Location

Target File: /worker/src/constants/default-model-prices.json

This JSON file contains an array of model pricing definitions used for cost calculation.

Data Structure

Complete Model Entry Schema

{
  "id": "uuid-generated-with-uuidgen",
  "modelName": "model-name-identifier",
  "matchPattern": "(?i)^regex-pattern$",
  "createdAt": "ISO-8601-timestamp",
  "updatedAt": "ISO-8601-timestamp",
  "tokenizerConfig": null,
  "tokenizerId": "claude|openai|null",
  "pricingTiers": [
    {
      "id": "model-uuid_tier_default",
      "name": "Standard",
      "isDefault": true,
      "priority": 0,
      "conditions": [],
      "prices": {
        "input": 0.000005,
        "output": 0.000025
      }
    }
  ]
}

Required Fields

Optional Fields

Pricing Tier Structure

Default Tier (Required)

Every model must have exactly one default tier:

{
  "id": "{model-id}_tier_default",
  "name": "Standard",
  "isDefault": true,
  "priority": 0,
  "conditions": [],
  "prices": { }
}

Rules for Default Tier:

• isDefault: Must be true
• priority: Must be 0
• conditions: Must be empty array []

Additional Tiers (Optional)

For usage-based pricing (e.g., large context pricing):

{
  "id": "uuid-for-tier",
  "name": "Large Context (>200K)",
  "isDefault": false,
  "priority": 1,
  "conditions": [
    {
      "usageDetailPattern": "(input|prompt|cached)",
      "operator": "gt",
      "value": 200000,
      "caseSensitive": false
    }
  ],
  "prices": { }
}

Condition Operators: gt, gte, lt, lte, eq, neq

Gathering Pricing Information

IMPORTANT: Always fetch pricing from official provider documentation. Never use heuristics or assumptions.

Official Pricing Sources

Required Information to Gather

1. Base input token price (per MTok)
2. Output token price (per MTok)
3. Cache write price (if caching supported)
4. Cache read price (if caching supported)
5. Long context pricing (if different tiers exist)
6. Model ID formats for all platforms (API, Bedrock, Vertex)

Price Conversion

Prices in the JSON are per token, not per million tokens.

Formula: price_per_token = price_per_mtok / 1_000_000 or price_per_mtok * 1e-6

Common Price Keys by Provider

Anthropic Claude Models

{
  "input": <base_input_price>,
  "input_tokens": <base_input_price>,
  "output": <output_price>,
  "output_tokens": <output_price>,
  "cache_creation_input_tokens": <cache_write_price>,
  "input_cache_creation": <cache_write_price>,
  "cache_read_input_tokens": <cache_read_price>,
  "input_cache_read": <cache_read_price>
}

OpenAI Models

{
  "input": <input_price>,
  "input_cached_tokens": <cached_input_price>,
  "input_cache_read": <cached_input_price>,
  "output": <output_price>
}

Google Gemini Models

{
  "input": <input_price>,
  "input_modality_1": <input_price>,
  "prompt_token_count": <input_price>,
  "promptTokenCount": <input_price>,
  "input_cached_tokens": <cached_price>,
  "cached_content_token_count": <cached_price>,
  "output": <output_price>,
  "output_modality_1": <output_price>,
  "candidates_token_count": <output_price>,
  "candidatesTokenCount": <output_price>
}

Match Pattern Examples

Anthropic Claude (API + Bedrock + Vertex)

(?i)^(anthropic\/)?(claude-opus-4-6|(eu\\.|us\\.|apac\\.)?anthropic\\.claude-opus-4-6-v1(:0)?|claude-opus-4-6)$

Matches:

• claude-opus-4-6 (direct API)
• anthropic/claude-opus-4-6 (with prefix)
• anthropic.claude-opus-4-6-v1:0 (AWS Bedrock)
• us.anthropic.claude-opus-4-6-v1:0 (regional Bedrock)
• claude-opus-4-6 (GCP Vertex)

With Version Date

(?i)^(anthropic\/)?(claude-opus-4-5-20251101|(eu\\.|us\\.|apac\\.)?anthropic\\.claude-opus-4-5-20251101-v1:0|claude-opus-4-5@20251101)$

OpenAI

(?i)^(openai\/)?(gpt-4o)$

Google Gemini

(?i)^(google\/)?(gemini-2.5-pro)$

Pattern Components

Step-by-Step: Adding a New Model

Step 1: Fetch Official Pricing

Use WebFetch to get pricing from official documentation:

WebFetch URL: https://platform.claude.com/docs/en/about-claude/pricing
Prompt: Extract pricing for [model name] including input, output, cache write, cache read prices per MTok

Step 2: Generate UUID

uuidgen
# Output: 13458BC0-1C20-44C2-8753-172F54B67647
# Convert to lowercase: 13458bc0-1c20-44c2-8753-172f54b67647

Step 3: Create the Entry

Example for a model with $5 input, $25 output, $6.25 cache write, $0.50 cache read:

{
  "id": "13458bc0-1c20-44c2-8753-172f54b67647",
  "modelName": "claude-opus-4-6",
  "matchPattern": "(?i)^(anthropic\/)?(claude-opus-4-6|(eu\\.|us\\.|apac\\.)?anthropic\\.claude-opus-4-6-v1(:0)?|claude-opus-4-6)$",
  "createdAt": "2026-02-09T00:00:00.000Z",
  "updatedAt": "2026-02-09T00:00:00.000Z",
  "tokenizerConfig": null,
  "tokenizerId": "claude",
  "pricingTiers": [
    {
      "id": "13458bc0-1c20-44c2-8753-172f54b67647_tier_default",
      "name": "Standard",
      "isDefault": true,
      "priority": 0,
      "conditions": [],
      "prices": {
        "input": 5e-6,
        "input_tokens": 5e-6,
        "output": 25e-6,
        "output_tokens": 25e-6,
        "cache_creation_input_tokens": 6.25e-6,
        "input_cache_creation": 6.25e-6,
        "cache_read_input_tokens": 0.5e-6,
        "input_cache_read": 0.5e-6
      }
    }
  ]
}

Step 4: Insert Entry

Add the entry to the JSON array in /worker/src/constants/default-model-prices.json.

Placement: Insert near related models (e.g., other Claude models together).

Step 5: Add to LLM Types (for Playground & LLM-as-Judge)

To make the model available in the Langfuse playground and for LLM-as-a-judge evaluations, add it to the appropriate model array in /packages/shared/src/server/llm/types.ts.

File: /packages/shared/src/server/llm/types.ts

Model Arrays by Provider:

• anthropicModels - Anthropic Claude models
• openAIModels - OpenAI GPT models
• vertexAIModels - Google Vertex AI models
• googleAIStudioModels - Google AI Studio models

IMPORTANT: Do NOT add new models as the first entry in the array. The first entry is used as the default model for test LLM API calls, and newer models may not be available to all users yet.

Example for Anthropic:

export const anthropicModels = [
  "claude-sonnet-4-5-20250929",  // Keep existing first entry
  "claude-haiku-4-5-20251001",
  "claude-opus-4-6",              // Add new model here (not first!)
  "claude-opus-4-5-20251101",
  // ... rest of models
] as const;

Step 6: Validate

# Check JSON syntax
jq . /path/to/default-model-prices.json > /dev/null && echo "Valid JSON"

# Verify entry exists
jq '.[] | select(.modelName == "claude-opus-4-6")' /path/to/default-model-prices.json

Multi-Tier Pricing Example

For models with long context pricing (e.g., different rates above 200K tokens):

{
  "id": "uuid-here",
  "modelName": "model-name",
  "matchPattern": "...",
  "pricingTiers": [
    {
      "id": "uuid-here_tier_default",
      "name": "Standard",
      "isDefault": true,
      "priority": 0,
      "conditions": [],
      "prices": {
        "input": 5e-6,
        "output": 25e-6
      }
    },
    {
      "id": "uuid-for-large-context-tier",
      "name": "Large Context (>200K)",
      "isDefault": false,
      "priority": 1,
      "conditions": [
        {
          "usageDetailPattern": "(input|prompt|cached)",
          "operator": "gt",
          "value": 200000,
          "caseSensitive": false
        }
      ],
      "prices": {
        "input": 10e-6,
        "output": 37.5e-6
      }
    }
  ]
}

Validation Rules

The system validates pricing tiers with these rules:

1. Exactly one default tier with isDefault: true
2. Default tier must have priority: 0 and empty conditions: []
3. Non-default tiers must have priority > 0 and at least one condition
4. All priorities must be unique within a model
5. All tier names must be unique within a model
6. Each tier must have at least one price
7. All tiers must have identical usage type keys
8. Regex patterns must be valid and safe (no catastrophic backtracking)

Common Mistakes

Using heuristics instead of official pricing:

// Wrong - assuming cache is 1.25x input
"cache_creation_input_tokens": input_price * 1.25

// Correct - use exact value from official docs
"cache_creation_input_tokens": 6.25e-6

Incorrect Price Format:

// Wrong - using MTok price directly
"input": 5

// Correct - price per token
"input": 5e-6

Missing Tier ID Suffix:

// Wrong
"id": "some-uuid"

// Correct for default tier
"id": "model-uuid_tier_default"

Invalid Regex Escaping:

// Wrong - unescaped dots
"matchPattern": "anthropic.claude"

// Correct - escaped dots
"matchPattern": "anthropic\\.claude"

Testing Model Matching

After adding a model, test that the regex matches expected inputs:

const pattern = new RegExp(matchPattern);
console.log(pattern.test("claude-opus-4-6")); // true
console.log(pattern.test("anthropic/claude-opus-4-6")); // true
console.log(pattern.test("anthropic.claude-opus-4-6-v1:0")); // true
console.log(pattern.test("us.anthropic.claude-opus-4-6-v1:0")); // true

Reference: Existing Model Entries

Look at these existing entries as templates:

Related Files

• Pricing JSON: /worker/src/constants/default-model-prices.json
• LLM Types: /packages/shared/src/server/llm/types.ts (model arrays for playground/LLM-as-judge)
• Validation: /packages/shared/src/server/pricing-tiers/validation.ts
• Matcher: /packages/shared/src/server/pricing-tiers/matcher.ts
• Tests: /web/src/__tests__/async/model-pricing-tiers.servertest.ts

Skill Status: COMPLETE Line Count: ~340 lines