Hyperbolic
AI cloud for renting NVIDIA H100, H200, B200 and B300 GPUs on demand or reserved, with managed Kubernetes and a pay-per-token OpenAI-compatible API for open text models.
About Hyperbolic
Getting GPU capacity for training, fine-tuning or inference usually means long procurement, upfront commitments and comparing many providers.
Who it’s for
- AI startups and researchers who need GPU instances in minutes without a sales process
- Developers who want to call open text models through an OpenAI-compatible API and pay per token
When to consider it
Consider Hyperbolic when you want self-serve hourly GPUs that can move to discounted reserved terms, plus hosted inference for open models, on one prepaid balance.
Tradeoffs & limitations
- Prepaid credits (minimum $5, never expire): personal accounts on the Free Tier cannot call inference until a one-time deposit of $5 or more upgrades them to Pro; on-demand GPUs are billed hourly and need a balance covering at least one hour, and reserved instances are paid up front and cannot be terminated early.
- Inference serves text models only (no image, speech or vision models) and is limited to 300 requests per minute per API key.
SOTA overview · Documentation-based assessment · Sources & review method
Updates
No updates shared yet.
Discussion
Newest firstAsk a question or share how you use Hyperbolic.
Keep it helpful. Community rules
Loading discussion…
Sources & review method
Documentation-based assessment · Nov 1, 2026 · Prepared with AI assistance; not a hands-on benchmark.
- Hyperbolic official website
- Hyperbolic docs overview
- Hyperbolic GPU pricing
- Hyperbolic inference overview
- Hyperbolic billing and payments
Checked by SOTA · AI-assisted documentation review. Selection advice is our assessment; verify current requirements for your deployment.
Import history & original evidence
Hyperbolic official website
AI cloud for renting NVIDIA H100, H200, B200 and B300 GPUs on demand or reserved, with managed Kubernetes and a pay-per-token OpenAI-compatible API for open text models.
Based on official pages and announcements checked on 2026-11-01. No hands-on test, performance benchmark or popularity ranking is claimed.