Product

From a usage export to a finance-ready margin view of your inference spend.

InfFyn computes GPp1M — Gross Profit per 1M Tokens — by model and by feature, then shows you where the margin is leaking, in dollars.

01

Connect read-only — or upload a CSV

Connect your model providers with read-only scopes. Or, if you'd rather not hand over credentials at all, export your usage to CSV and upload it. The audit runs either way.

  • Read-only API access — usage metadata only
  • CSV upload path for the security-conscious — no credentials needed
  • Five-minute setup, no engineering work

Option A

Read-only connect

OpenAI · Anthropic · Google · Together · self-hosted

openai.comREAD-ONLY
anthropic.comREAD-ONLY
google aiREAD-ONLY

Option B

CSV upload

No credentials. No third-party access.

usage-export.csv

drag and drop

02

See your spend per 1M tokens, by model

The free audit returns spend per 1M tokens by model so you can see which segments cost far more than others, plus one quantified waste finding — ranked by dollar impact. Add your revenue in the detailed audit to turn spend into GPp1M — profit per 1M tokens by segment.

  • Spend per 1M tokens by model and by feature
  • One waste finding free — the biggest dollar leak we found
  • Upgrade: add revenue to compute GPp1M (profit) by segment

Top finding

Classifier traffic running on gpt-4o

−$8,420/mo

610K requests/month are routine classification calls hitting gpt-4o. A smaller model would handle this segment at a fraction of the per-token cost.

GPp1M today

−$41

Est. GPp1M

+$118

Illustrative — your audit uses your data.

03

Turn the snapshot into continuous monitoring

Inference spend drifts. A new prompt rolls out, traffic shifts, a model price changes — and your margin moves with it. InfFyn watches continuously and alerts you when GPp1M drifts on any segment.

  • Drift alerts when GPp1M moves on a model or feature
  • Monthly margin and run-rate-waste report your CFO can actually read
  • Forecast view that ties inference spend to revenue

Drift alert · just now

MARGIN ↓

gpt-4o (router) GPp1M dropped from +$22 to −$118 over the last 7 days.

Likely cause: prompt change shipped Tuesday

−$3.2K projected/mo

CFO view

The same data, reframed as gross margin.

Engineering sees per-model GPp1M and drift. Finance sees gross margin impact, monthly run-rate waste, and a forecast. One number, two languages.

Gross margin Δ

+3.4 pts

Run-rate waste

$104K/yr

Forecast Q4

$1.21M

Exports as a one-page PDF for your board deck or month-end close.

Security

Built so security teams don't block it.

Read-only scopes

We request the minimum scope needed to read usage metadata. Never write, never billing.

No raw-log storage

We compute on usage counters and metadata — not the contents of your prompts or completions.

Retention policy

Uploaded CSVs are processed and deleted on a defined retention window. Audit results stay in your account.

See your spend per 1M tokens in about five minutes.