How does the LLM model migration calculator work?

Select your current model, enter your daily token usage and input/output ratio, and optionally set a minimum quality threshold. The tool calculates your current monthly cost using verified pricing, then compares every alternative model — showing monthly savings, cost difference, and quality tradeoff based on benchmark data.

What is performance per dollar (perf/$)?

Performance per dollar is a model's average benchmark score divided by its blended cost per million tokens. Higher is better — it means you get more benchmark performance per dollar spent. A model with 60 benchmark score at $2/Mtok blended has perf/$ of 30, while a model with 80 score at $15/Mtok has perf/$ of 5.3.

Should I always switch to the cheapest model?

No. The cheapest model may not meet your quality requirements. Use the quality threshold slider to filter out models that would degrade output quality below your acceptable level. The best migration target is usually a model that maintains 90%+ of your current quality while significantly reducing cost — often a mid-tier model rather than the absolute cheapest.

Pricing / Migration Calculator

LLM Model Migration Calculator

See exactly how much you'd save by switching from your current model to any alternative — with quality tradeoff analysis from real benchmark data.

⚡ Your Current Setup

Current model the model you're using now

Tokens per day ~1M tokens ≈ 750K words

Input / Output ratio 70% input

Days per month operating days

Minimum quality threshold Show all models

0 (no filter) 100 (best only)

📊 Your Current Costs

Select your current model to begin.

Migration targets

All alternatives ranked by savings

Select your current model above to see migration options.

How the migration calculator works

This tool combines two datasets: verified API pricing (from official provider pages, updated 3× daily) and benchmark performance data (from the Vellum LLM Leaderboard, scraped daily). When you select your current model, we calculate your exact monthly cost using the input/output token ratio you specify.

For each alternative model, we show:

Monthly cost — your projected spend on that model, using your same token usage
Monthly savings — how much less (or more) you'd pay vs your current model
Quality — the model's average benchmark score (higher = better output quality)
Perf/$ — performance per dollar: benchmark score ÷ blended cost per Mtok (higher = better value)
Verdict — a quick assessment: "save X%, quality Y% better/worse"

Quality threshold: Use the slider to filter out models below your quality floor. If your current model scores 80 on benchmarks, setting the threshold to 70 will only show alternatives with 70+ scores — ensuring you don't sacrifice too much quality for cost savings.

Important: Benchmark scores are averages across multiple tasks and may not reflect your specific use case. A model with a lower average score might outperform a higher-scored model on your particular workload. Use this tool as a starting point, then test promising candidates on your actual data.