Right now, every API call you make is priced at the floating rate — whatever the provider charges today. That rate moves. Sometimes in your favor (price cuts), sometimes against you (new premium models, tier changes). You can't budget against a moving target.

A fixed rate is an agreed price that doesn't change for the duration of a contract. You know exactly what you'll pay per million tokens next month, next quarter, regardless of what the market does.

EICI is a weighted benchmark that tracks the average cost of AI inference across 42 models and 7 providers. It's the floating rate — a single number that tells you what the market charges today.

When OpenAI cuts prices, when Anthropic launches a new tier, when Google adjusts Gemini — EICI reflects it. Updated daily. It's the reference price both sides of a swap agree on.

A swap is an agreement to exchange a floating price for a fixed one. With EICS, you swap your unpredictable per-token cost for a known fixed rate, settled against the EICI benchmark.

You choose a term (1M, 3M, or 6M). Exostream prices the swap using the EICI forward curve — a projection of where inference costs are headed based on historical decay rates. The result is a fair fixed rate for that period.

If the market moves above your fixed rate, you're protected. If it drops below, you're overpaying — but you bought predictability. That's the tradeoff.

Inference costs have been falling. So why would you lock in a fixed rate on something that's getting cheaper? Because the trend isn't guaranteed, and even if it holds, the rate and timing of decline are unpredictable.

Track what you're actually spending across models.

κ tells you how exposed you are to price changes — high context = high exposure.

Forward curves show where prices are headed so you can plan migrations.

Cost alerts when a cheaper model crosses below your current one.

Budget forecasting with forward curves, not guesswork.

The oracle publishes θ (decay rate) — how fast prices are falling per model.

Inference Budget Planner: input your monthly volume, get spot + forward cost projections.

"Inference costs fell 4.2% last month per the Exostream Index" — the number for your CFO.

Replace hardcoded price tables with a live API feed.

Observability platforms: accurate auto-updating cost tracking.

Router/gateway projects: real-time pricing for cost-aware routing decisions.

One integration, all providers, always current.

Structured historical pricing data that doesn't exist anywhere else.

Depreciation analytics: half-life of pricing power per model generation.

Provider competitive intelligence: who's cutting fastest, where margins compress.

The EICI — a single benchmark for the market.

The Cost Calculator prices a single API call. Open it from the Calculators page.

The EPM generates forward curves using theta-based exponential decay fitted to each model family's historical pricing. We backtest these projections against realised prices to measure model fit.

Important context: Forward curves in any market are not predictions — they are a consistent, reproducible basis for pricing future contracts. Interest rate forwards are routinely "wrong" about where rates end up. Oil futures miss by 30-50% over 6 months. The value is in providing a transparent, arbitrage-consistent reference that counterparties can agree on, not in forecasting accuracy.

AI inference pricing is especially volatile: providers make discrete strategic pricing decisions (50% overnight cuts, successor models launching at higher price points) that no smooth decay model can anticipate. The theta-based curve is a bootstrap mechanism — as swap market liquidity develops, the market itself will determine the forward curve through trading.

GET https://api.exostream.ai/v1/backtest

# Returns: tenor summaries, family breakdowns,
# individual predictions with predicted vs actual prices,
# filtering metadata (audit corrections, step-function events)

How Exostream ensures pricing data quality and reliability.