Huggingface Removes TPU v5e Hardware Options, Impacting Pricing for Spaces
What happened
Between the Feb 1, 2025 and Apr 11, 2025 captures of Huggingface's pricing page, Pulse recorded 4 changes: TPU v5e 2x2 hardware option ($4.75/hr) removed from Spaces Hardware; Pro Account: 'Higher rate limits for serverless inference' replaced with 'Free credits across all Inference Providers' ($2 credits included); TPU v5e 1x1 hardware option ($1.20/hr) removed from Spaces Hardware; TPU v5e 2x4 hardware option ($9.50/hr) removed from Spaces Hardware. Pulse files it as a packaging change (limit raised, feature change). Pulse's latest capture (Nov 15, 2026) lists PRO at $9, Team at $20, Enterprise at $50, Inference Endpoints - NVIDIA T4 AWS (4x) at $3 per hour and Inference Endpoints - NVIDIA RTX PRO 6000 AWS (1x) at $2.75 per hour.
Huggingface's previous recorded change was "Huggingface Reduces GCP NVIDIA A100 Inference Endpoint Prices by 40% to $3.60/hour, Delivering Significant Cost Savings" (Q4 2024). Huggingface has been tracked since 2023-11-14.
In the same window Cohere, OpenAI and Anthropic also changed their AI & Machine Learning pricing pages. Those moves are listed further down, and the Pulse Agent can compare them side by side.
- Removed—Limit
BeforeTPU v5e 2x2 hardware option ($4.75/hr)
AfterNot listed
- Feature changePROFeature
Pro Account: 'Higher rate limits for serverless inference' replaced with 'Free credits across all Inference Providers' ($2 credits included)
- Removed—Limit
BeforeTPU v5e 1x1 hardware option ($1.20/hr)
AfterNot listed
- Removed—Limit
BeforeTPU v5e 2x4 hardware option ($9.50/hr)
AfterNot listed
| Plan | Field | Before | After | Change |
|---|---|---|---|---|
| — | Limit | TPU v5e 2x2 hardware option ($4.75/hr) | Not listed | Removed |
| PRO | Feature | Pro Account: 'Higher rate limits for serverless inference' replaced with 'Free credits across all Inference Providers' ($2 credits included) | Feature change | |
| — | Limit | TPU v5e 1x1 hardware option ($1.20/hr) | Not listed | Removed |
| — | Limit | TPU v5e 2x4 hardware option ($9.50/hr) | Not listed | Removed |
Values as written on Huggingface's public pricing page in the two captures Pulse compared; where the page states no number, the cells give the wording before and after.
Evidence
The two captures Pulse comparedFull-page captures of Huggingface's pricing page (huggingface.co) from Feb 1, 2025 and Apr 11, 2025; the change is marked around the middle of the page, where the TPU v5e 2x2 hardware option ($4.75/hr) no longer appears.
Sign in to see the evidence
The two full-page captures with this change marked up are free with an account.
Huggingface, the last 90 days
Full history- Jan 14, 2026PackagingHuggingface Discontinues Key Hardware Options and Persistent Storage Tiers, Impacting User Access and Pricing
- Q3 2025PackagingHuggingface Introduces Cost-Effective AWS NVIDIA H100 Inference Endpoints at $4.50/hour, Reducing Prices from GCP's $10.00/hour
- Q2 2025PackagingHuggingface Introduces New Team Plan and Upgrades PRO Features Amid Significant Price Decrease for A100 Hardware
- Q1 2025Limit raisedHuggingface Removes TPU v5e Hardware Options, Impacting Pricing for Spaces This change
- Q4 2024Price decreaseHuggingface Reduces GCP NVIDIA A100 Inference Endpoint Prices by 40% to $3.60/hour, Delivering Significant Cost Savings
Across AI & Machine Learning in the same window
All AI & Machine Learning changes- CohereJan 13Cohere Introduces Command R7B Model, Optimizing Speed and Efficiency at Competitive Pricing
- OpenAIJan 12OpenAI Announces Dramatic 87% Price Drop for Flagship Reasoning Model Outputs During Transition to o3
- AnthropicJan 10Anthropic API Introduces Tiered Pricing and Prompt Caching for Sonnet 4.5, Restructuring Costs Based on Token Length
Questions about this change
Did Huggingface raise prices in April 2025?
No list-price increase was recorded between Feb 1, 2025 and Apr 11, 2025. The change Pulse detected was a packaging change: limit raised, feature change.
What changed on Huggingface's pricing page between Feb 1, 2025 and Apr 11, 2025?
Limit removed: TPU v5e 2x2 hardware option ($4.75/hr); Pro Account: 'Higher rate limits for serverless inference' replaced with 'Free credits across all Inference Providers' ($2 credits included); limit removed: TPU v5e 1x1 hardware option ($1.20/hr); limit removed: TPU v5e 2x4 hardware option ($9.50/hr). Pulse compared the captures it took on Feb 1, 2025 and Apr 11, 2025; the before/after table above lists each item as written on the page.
How often does Huggingface change its pricing?
Pulse recorded 1 other pricing-page change for Huggingface in the 90 days before this one; the previous change was "Huggingface Reduces GCP NVIDIA A100 Inference Endpoint Prices by 40% to $3.60/hour, Delivering Significant Cost Savings" in Q4 2024; Huggingface has been tracked since 2023-11-14.