MODE VC
AI Economics Workbench 06

LLM Inference Cost &
Unit Economics Calculator

Model monthly token consumption, API costs, prompt caching economics, and gross margin impact across Claude 3.7, GPT-4o, DeepSeek V4.1, Gemini 2.5, and open-source models. Runs completely client-side in your browser.

Workload Parameters

Saves up to 75% to 90% on repeated system prompts and context.

Monthly LLM Cost $1,452 Prompt cache saved $360
Cost / User / Mo $0.29 $0.0073 / request
Gross Margin 98.5% Excellent unit economics
Total Monthly Requests: 200,000
Total Monthly Input Tokens: 240.0M
Total Monthly Output Tokens: 80.0M

Cross-Model Economics Matrix

Monthly spend under this exact workload across frontier and distilled models.

Model In / Out ($/1M) Monthly Spend Gross Margin
Chart • Gross margin by model as usage per user grows

Gross margin by model as requests per user grow.

Each panel is one model at your price, tokens and cache rate. The dot is your current requests per user; the line runs from 10 to 640 requests a user a month.

Same pricing table as the matrix above. Calculated in your browser.Drawn with Vega-Lite
Interactive model ARPU vs. LLM Spend
Subscription ARPU vs. Monthly LLM Cost
Click or use Left/Right arrows to compare monthly revenue against token inference spend.

Tri-Tier Routing Advantage

Recommended

By routing 80% of routine workflows (classification, parsing, extraction) to high-speed distilled models (DeepSeek V4.1 or Gemini Flash) and reserving frontier reasoning (Claude 3.7 or GPT-4o) for the top 20% complex cases, monthly LLM cost reduces to $362 (a 75.1% burn reduction).

Hardware speed benchmarks: pilot in progress

We withdrew our earlier throughput and latency figures and are re-measuring from three regions with paid keys. No speed numbers are shown until that run is published. See the planned method →

Explore empirical price trends: AI Inference Cost Comparison Chart →