`cost` — Token Count & Cost Analysis

Severity: INFO · Auto-fix: No · Category: 💰 Cost

What It Does

Reports the prompt's token count and estimated cost per API call. This is always informational — it never causes a pipeline failure on its own.

Example Output

[ INFO ] cost (line -)
  Prompt is ~97 tokens (~$0.0005 input per call on gpt-4o).
  At 10,000 calls/day → ~$4.85/day input.

Cost Dashboard

Run with --show-dashboard for a full savings breakdown:

bash

promptlint --file prompt.txt --show-dashboard

  Savings Dashboard
  ──────────────────────────────────────────────────
  Current Tokens:    97
  Optimized Tokens:  59  (39.2% reduction)

  Savings per Call:  ~$0.0002
  Daily Savings:     ~$2.00 at 10,000 calls/day
  Monthly Savings:   ~$60.00
  Annual Savings:    ~$720.00

Configuration

yaml

model: gpt-4o
cost_per_1k_tokens: 0.005
calls_per_day: 10000

rules:
  cost: true  # or false to disable

Token Counting

The Python CLI uses tiktoken for exact counts when installed:

bash

pip install "promptlint-cli[tiktoken]"

Without tiktoken (or in the Node.js CLI), tokens are estimated as len(text) / 4 (±15%).

Common Model Prices

Model	Input per 1K tokens
GPT-4o	$0.005
GPT-4o mini	$0.00015
GPT-4	$0.030
GPT-3.5 Turbo	$0.0015
Claude 3.5 Sonnet	$0.003
Claude 3 Haiku	$0.00025
Gemini 1.5 Pro	$0.00125

Relationship to `cost-limit`

The cost rule is purely informational. If you want to enforce a token budget, use cost-limit.

cost — Token Count & Cost Analysis ​

What It Does ​

Example Output ​

Cost Dashboard ​

Configuration ​

Token Counting ​

Common Model Prices ​

Relationship to cost-limit ​