Today we're adding Hugging Face to the Stack Spend provider panel. If you use Hugging Face for Inference Endpoints, Spaces, or Jobs, you can now track that spend in the same dashboard as OpenAI, Anthropic, Cursor, and your cloud providers. One place for closed-source and open-source AI costs.
Why Hugging Face Matters for AI Cost Tracking
Hugging Face is the central platform for open-source AI. While OpenAI and Anthropic dominate closed-source models, Hugging Face hosts hundreds of thousands of open models—Llama, Mistral, Qwen, Phi, and many more. Enterprises use Hugging Face for Inference Endpoints (dedicated model serving), Spaces (hosted apps), and Jobs (batch inference). Those services incur real costs. Until now, they've been separate from your main AI cost view.
Adding Hugging Face to Stack Spend changes that. You get:
- Unified AI spend—OpenAI, Anthropic, Cursor, and Hugging Face in one dashboard
- Provider-level visibility—see how much you're spending on open vs closed models
- Same forecasting and alerts—anomaly detection, daily monitoring, and pace-to-forecast for Hugging Face just like your other providers
- Dramatically more models tracked—from a handful of closed-source APIs to hundreds of open models via Hugging Face
Hugging Face's Positioning: Open Models Alongside Closed
Hugging Face positions itself as the alternative to proprietary AI. Where OpenAI and Anthropic offer closed APIs with opaque pricing, Hugging Face offers open models you can run yourself—on their infrastructure or yours. Enterprises use it for:
- Cost control—open models can be cheaper than GPT-4 or Claude at scale
- Data privacy—models run in your chosen region; data stays where you configure it
- Model choice—pick the right model for the task instead of being locked into one vendor
- Transparency—open weights, open benchmarks, no black box
Hugging Face has partnerships with AWS, Google Cloud, and Azure. Their Enterprise Hub offers SSO, audit logs, and data residency controls. The platform has evolved from a model repository into a full inference and deployment layer. That layer has costs. Stack Spend now tracks them.
What Stack Spend Tracks for Hugging Face
When you connect your Hugging Face organization, Stack Spend pulls usage and cost data for:
- Inference Endpoints—dedicated model serving (CPU and GPU instances, billed per hour)
- Spaces—hosted ML applications (Gradio, Streamlit, static sites)
- Jobs—batch inference and training
- Inference Providers—serverless inference
- Private Storage—model and dataset storage
Costs are normalized to USD and rolled up daily, so you get the same trend views, forecasts, and anomaly detection as for OpenAI or Anthropic. You can filter by service (Endpoints vs Spaces vs Jobs) and see how open-source model spend compares to closed-source.
Closing the Gap Between Closed and Open
Most teams track OpenAI and Anthropic. Fewer track Hugging Face. The result is a gap: you know closed-source spend, but not open-source. As more teams adopt Llama, Mistral, and other open models via Hugging Face, that gap becomes a blind spot.
Stack Spend now closes it. Connect Hugging Face alongside OpenAI, Anthropic, Cursor, AWS, GCP, and Azure. See total AI spend—closed and open—in one place. Forecast it. Alert on it. No more spreadsheets, no more delayed reconciliation.
To get started: Add Hugging Face from the Providers panel. You'll need an access token and your organization slug. Our Hugging Face setup guide walks through the steps.