Track Fireworks AI spend in Cloud Intelligence™

Connect your Fireworks AI account and get cost and usage analytics, budgets and attribution by model, deployment type, and GPU, next to the rest of your cloud and AI bills.

AI inference spend has a habit of living outside your cloud bill. If you run models on Fireworks AI, you're paying per token for serverless inference, per GPU-hour for dedicated deployments, and separately for fine-tuning, and none of it shows up where your FinOps work happens. Checking the Fireworks dashboard is a manual job, and "which model is driving this month's increase" has no good answer once more than one team holds an API key.

That gap is closed. Cloud Intelligence™ now imports Fireworks AI cost and usage data.

What you get

Once connected, Fireworks AI shows up as a provider in Cost & Usage reports. Costs arrive at daily granularity, grouped by product line (Serverless Inference, Dedicated Deployments, Fine-tuning) and by model, with usage measured in tokens for inference and fine-tuning and in GPU-hours for dedicated deployments. Batch inference is broken out too, including whether input tokens were served from cache.

The auto-generated labels do the heavy lifting: Model, Deployment Type, Usage Type, Cached, and Accelerator (the GPU type behind a dedicated deployment). If your dedicated deployment idles on an 8-GPU node all weekend, the report says so in GPU-hours, not in a surprise at month end.

On first connection we backfill up to 180 days of history, then refresh every six hours. Fireworks AI spend also appears in the AI Intelligence dashboard alongside your Anthropic, OpenAI, Cursor, and other AI provider costs, so total AI spend is one view, not five tabs.

Setup is deliberately boring. You need your Fireworks Account ID and an admin API key, and the connection test tells you immediately whether they work. The Fireworks AI connection guide has the exact steps.

Get started

1. Connect your Fireworks AI account from the Integrations catalog.
2. Wait for the initial import (you'll get an email when it's done, usually within 30 minutes).
3. Navigate to AI Intelligence dashboard to see your Fireworks AI spend alongside other AI providers.

If you're running real inference workloads on Fireworks AI, we'd like to hear what the reports do and don't answer for you — that feedback shapes what we import next.