What inputs does the Model Price Drop Stress Test need?

It takes 3 inputs: monthlyRevenue, monthlyAiCost, grossMarginPercentToday. Outputs returned: scenarios, mostLikelyMargin.

Can I verify the Model Price Drop Stress Test with a worked example?

Yes. With $100k revenue, $15k AI, 60% margin today. the tool returns 50%-drop keep-savings margin ≈ 67.5%.

Where does the Model Price Drop Stress Test get its benchmark data?

Reference data is sourced from: Anthropic API pricing history (as of 2026-04); OpenAI API pricing history (as of 2026-04); a16z LLM token price trends 2024–2025 (as of 2025-Q4).

What can the Model Price Drop Stress Test not tell me?

Known limitations: Doesn't model multi-vendor stacks — treat the combined AI line as a single field. Reality is between keep-savings and pass-through; the model gives bounds, not point estimates.

Methodology: Model Price Drop Stress Test

1. Scope

Recomputes gross margin under 10/30/50% LLM price drop scenarios under two narratives: keep-savings and full pass-through to customers.

2. Inputs and outputs

Inputs

monthlyRevenue number ($)
monthlyAiCost number ($)
grossMarginPercentToday number (0–100)

Outputs

scenarios

10/30/50% drop, both keep-savings and pass-through margins.
mostLikelyMargin

50% keep-savings margin (best-case anchor).

Engine source: src/lib/model-price-drop-stress-test/engine.ts

3. Formula / scoring logic

today_cogs = revenue − revenue × margin/100
non_ai_cost = today_cogs − ai_cost
for each drop:
  new_ai = ai × (1 − drop/100)
  new_margin_keep = (revenue − non_ai − new_ai) / revenue × 100
  rev_pass = revenue × (1 − ai_share × drop/100)
  new_margin_pass = (rev_pass − non_ai − new_ai) / rev_pass × 100

4. Assumptions

AI cost is fully variable. Reserved or committed spend doesn't drop with list prices.
Pass-through compresses revenue by AI's share of original COGS — a reasonable mid-case for competitive markets.

5. Data sources

Anthropic API pricing history as of 2026-04
OpenAI API pricing history as of 2026-04
a16z LLM token price trends 2024–2025 as of 2025-Q4

6. Known limitations

Doesn't model multi-vendor stacks — treat the combined AI line as a single field.
Reality is between keep-savings and pass-through; the model gives bounds, not point estimates.

7. Reproducibility

Input
$100k revenue, $15k AI, 60% margin today.

Expected output
50%-drop keep-savings margin ≈ 67.5%.

8. Change log

2026-05-08 methodology first published.

Worked example

Run live against the same engine this site ships (/engines/model-price-drop-stress-test.js). The inputs and outputs below are recomputed on every build and independently re-verified in CI — they are never hand-authored.

Input

tool: model_price_drop_stress_test
monthly_revenue: 100000
monthly_ai_cost: 15000
gross_margin_percent_today: 60

Output

todayGrossProfit: 60000
todayNonAiCost: 25000
scenarios[0].dropPercent: 10
scenarios[0].newAiCost: 13500
scenarios[0].newGrossMarginKeepSavings: 61.5
scenarios[0].newGrossMarginPassThrough: 60
scenarios[0].newRevenueIfPassThrough: 96250
scenarios[1].dropPercent: 30
scenarios[1].newAiCost: 10500
scenarios[1].newGrossMarginKeepSavings: 64.5
scenarios[1].newGrossMarginPassThrough: 60
scenarios[1].newRevenueIfPassThrough: 88750
scenarios[2].dropPercent: 50
scenarios[2].newAiCost: 7500
scenarios[2].newGrossMarginKeepSavings: 67.5
scenarios[2].newGrossMarginPassThrough: 60
scenarios[2].newRevenueIfPassThrough: 81250
mostLikelyMargin: 67.5

Frequently asked questions

What does the Model Price Drop Stress Test calculate?: Recomputes gross margin under 10/30/50% LLM price drop scenarios under two narratives: keep-savings and full pass-through to customers.
What inputs does the Model Price Drop Stress Test need?: It takes 3 inputs: monthlyRevenue, monthlyAiCost, grossMarginPercentToday. Outputs returned: scenarios, mostLikelyMargin.
What formula does the Model Price Drop Stress Test use?: The exact computation is: today_cogs = revenue − revenue × margin/100; non_ai_cost = today_cogs − ai_cost; for each drop:; new_ai = ai × (1 − drop/100); new_margin_keep = (revenue − non_ai − new_ai) / revenue × 100; rev_pass = revenue × (1 − ai_share × drop/100); new_margin_pass = (rev_pass − non_ai − new_ai) / rev_pass × 100
Can I verify the Model Price Drop Stress Test with a worked example?: Yes. With $100k revenue, $15k AI, 60% margin today. the tool returns 50%-drop keep-savings margin ≈ 67.5%.
Where does the Model Price Drop Stress Test get its benchmark data?: Reference data is sourced from: Anthropic API pricing history (as of 2026-04); OpenAI API pricing history (as of 2026-04); a16z LLM token price trends 2024–2025 (as of 2025-Q4).
What can the Model Price Drop Stress Test not tell me?: Known limitations: Doesn't model multi-vendor stacks — treat the combined AI line as a single field. Reality is between keep-savings and pass-through; the model gives bounds, not point estimates.