Model Economics
Which model actually earns its keep.
Compare cost and revenue, per model.
Per-model margin analytics built on maintained rate cards. Bear Lumen shows which models generate margin and which erode it, simulates provider changes against your real usage, and gives you the cost-per-request breakdown to back every model decision.
Margin per modelMaintained rate cardsBuilt with SOC 2 standards in mind
Every model looks the same on the invoice.
Provider bills report total API spend, not model economics. Teams pick between GPT-4o, Claude, and Gemini on benchmark vibes while the margin consequences of each choice stay invisible until month end.
- Provider invoices lump every model into one line
- No cost-per-request visibility by model or feature
- Margin impact of a model swap is anyone’s guess
- Provider price changes land silently in next month’s bill
Margin per model, side by side.
Every tracked request is attributed to its model. Bear Lumen maps maintained rate cards to that usage, then pairs cost with the revenue it serves, so each model carries its own cost-per-request, margin, and trend.
- Cost per request for every model you run
- Revenue minus cost-to-serve, broken down by model
- Maintained rate cards track provider price changes
- Provider rollups: OpenAI vs Anthropic vs Bedrock spend
Key Capabilities
Model economics, measured not guessed.
Every model you run gets the same treatment: cost, revenue, margin, and trend, computed from tracked usage and maintained rate cards.
Per-model margin analytics
Revenue minus cost-to-serveRevenue, cost, and margin for each AI model you use, queryable straight from the SDK. See which models generate margin, which erode it, and how each one trends as provider pricing moves.
- Break down by
- model, provider
- Metrics
- cost per request, margin, trend
- SDK surface
- bear.costs.byModel() · bear.margins.summary()
What-if simulation
Before/after margin impactModel a provider price increase or a switch from GPT-4o to Claude Sonnet. The simulation runs the change against your real usage data and shows projected margin impact per customer, before you change a line of code. Decide whether to absorb, pass through, or switch models.
Provider-level rollups
bear.costs.byProvider()Aggregate economics by provider: total OpenAI spend vs Anthropic vs Bedrock. Usage is normalized across every provider the SDK auto-detects, so the rollups are ready for vendor negotiations and budget planning.
Business Impact
Choose models on margin, not vibes.
Swap models with confidence
Compare cost-per-request and margin percentage side by side. Identify exactly where a model swap improves unit economics, measured against your real traffic.
Catch price changes early
Track how model costs move over time. Spot when a provider price drop creates an opportunity, or when an increase threatens the margins of a flagship feature.
Negotiate from data
Provider rollups give you the full spend picture per vendor. Walk into renewal conversations knowing what each provider’s models contribute to your margin.
Make model decisions with real data.
Per-model costs, margin analytics, and what-if simulation. Stop guessing which model is cheapest. Measure it.
No credit card required. Free tier included. Set up in minutes.