Hugging Face

Hugging Face is the Hub plus compute. Rechecked 2026-08-17: huggingface.co/pricing and huggingface.co/pro sell PRO $9 / month (20x included inference credits, PRO badge). Team is $20 (instant org setup; members get ZeroGPU and Inference Providers PRO benefits). Enterprise starts at $50 on the public compare (SCIM, legal/compliance; Enterprise Support is a separate row). Billing docs say extra private storage is $18 / TB / month above the included 1TB (or 1TB per seat). Compute (Inference Providers, Spaces GPUs) is billed separately. github.com/huggingface/transformers had 164,164 stars, license Apache-2.0. Drop "100,000 models / 10,000 datasets" as an audited census.

Compare Ollama if you wanted a local runner with Free / Pro $20 / Team $25, and OpenRouter if you wanted a multi-provider router.

Key Features

• Hub: Host models, datasets, and Spaces. The live catalog is the site search, not a 2024 count.
• PRO vs Team vs Enterprise: PRO is an individual subscription. Team/Enterprise are org subscriptions. Compute credits are extra.
• Transformers library: Apache-2.0. That is not the same as a PRO seat.
• Pay-as-you-go hardware: Pricing still lists on-demand GPUs starting near $0.03 / hour. Recheck the live hardware table.

Use Cases

• People who will pay PRO $9 for extra inference credits and a badge.
• Orgs that will pay Team $20 for centralized tokens.
• People who only need the OSS library: stay on Free Hub + Transformers.

Limitation: We did not re-count models or datasets. China-usage anecdotes are out.

Pricing

huggingface.co/pricing and billing on 2026-08-17.

If a leftover blog still lists only "Transformers is free GitHub," treat this table as the source for Hub seats.

Getting Started

1. Open huggingface.co/pricing before you buy GPU time.
2. Use Transformers from GitHub if you do not need hosted inference.
3. Recheck Ollama if you wanted weights on your own machine.
4. Do not paste 100k models into a deck.

Frequently Asked Questions

Is Hugging Face still just a free model zoo?

The Hub has a free tier. PRO is $9. Team is $20. Compute is extra.

Does PRO include unlimited GPUs?

No. PRO includes inference credits. Hardware is pay-as-you-go.

Same as OpenRouter?

No. Hugging Face is Hub plus its own inference/compute. OpenRouter routes many providers.

Alternatives

• Ollama: Local runner plus Ollama cloud.
• OpenRouter: Multi-provider API wallet.

Tips

1. Quote PRO $9 / Team $20 / extra storage $18/TB.
2. Keep Transformers stars separate from Hub seats.
3. Recheck Enterprise $50 before you treat it as a full contract.

Conclusion

Hugging Face is a priced Hub plus Apache-2.0 libraries, not a 100k-model slogan. Start at huggingface.co/pricing, then decide whether Ollama already covers the local models you need.