The Free Tier Math

Terse's free tier includes 1,500 prompt optimizations per week at $0. Run the numbers and that allowance is worth roughly $6 per week — about $310 per year — in avoided token spend for a working developer. The free tier alone often saves more than most paid AI tools cost.

What One Optimization Is Worth

"Free forever, 1,500 optimizations a week" is easy to skim past, so let's make it concrete. A typical prompt that Terse cleans up drops by 30–40% of its tokens — filler removed, whitespace collapsed, verbose phrases shortened, with the meaning intact (the same pipeline described in what is token optimization).

On a 2,000-token prompt, a 35% reduction saves about 700 tokens. At $3 per million input tokens — a common mid-tier API rate — that single optimization is worth about $0.002. Trivial in isolation. But nobody sends one prompt.

Multiplying by Real Usage

A developer working with an AI assistant all day sends a lot of prompts, and in agent sessions each optimized prompt also shrinks the conversation history that gets re-billed on every later turn. Working through the arithmetic for a heavy user who actually consumes the weekly allowance:

1,500 optimizations/week
× ~700 tokens saved each (input + re-billed history effect)
≈ 1M+ tokens of avoided spend per week
≈ $6/week at typical rates
≈ $310/year — on the free tier alone

Moderate users save proportionally less — someone using a third of the allowance still clears about $100 a year — but the shape of the result holds: the savings are a real budget line, not a rounding error. You can sanity-check the numbers against your own prompt sizes with the token calculator.

Why Give That Away Free?

Because the savings sell the upgrade better than any pitch could. Terse's Statistics view turns token reductions into dollar figures, per source and per day. Once a developer watches a week of savings accumulate — browser prompts, agent sessions, pasted diffs — the paid tier (unlimited optimizations, team stats, priority models) tends to justify itself in days. We would rather show the value than argue for it. The real-world results post shares what those dashboards look like for actual users after 30 days.

How to Run the Experiment Yourself

• Baseline first. Note this month's AI invoice before installing anything. You'll want the before number.
• Run Terse free for seven days. Change nothing about how you work — Normal mode handles filler and whitespace automatically.
• Check Statistics on day seven. The dollar total is computed from your actual prompts at your actual model rates, not from our averages.
• Compare to your invoice. For most daily users the savings line is a double-digit percentage of spend — consistent with the broader strategies in reducing AI API costs.
• Upgrade only if you cap out. Hitting 1,500 optimizations a week regularly is the signal that the paid tier already pays for itself. Until then, the free tier is the rational choice — we've done that math too.

Frequently Asked Questions

How much does the Terse free tier actually save?

A heavy user who consumes the full 1,500 weekly optimizations saves roughly $6/week — about $310/year — based on an average 30-40% token reduction per prompt at typical API rates. Moderate users save proportionally less but still a meaningful amount.

Is 1,500 optimizations a week enough for daily use?

For most individual developers, yes. It covers roughly 200-300 optimizations per working day across browser, editor, and agent sources. Users who consistently hit the cap are exactly the users for whom the paid tier is already cheaper than the waste.

Do the savings depend on which AI model I use?

The percentage reduction is model-independent — filler is filler. The dollar value scales with your model's input price: higher-priced frontier models make each saved token worth more, cheaper models less.

Further Reading

• Terse Blog — all token optimization guides
• Real Token Optimization Results — anonymized user data
• Why Optimize Tokens? — the full case
• Token Calculator — price your own prompts