Hyperbolic (hyperbolic.ai) is an AI cloud that sells on-demand, reserved and private GPU capacity and, separately, a serverless inference API described as OpenAI-compatible at base URL https://api.hyperbolic.xyz/v1 with 25+ open-source models. Its official rate-limit table defines a Basic tier of 60 requests per minute available to a free account with a $0 minimum deposit, and a Pro tier of 600 requests per minute that unlocks automatically once $5 or more is deposited. Serverless inference is pay-as-you-go: text generation runs from $0.10 per 1M tokens for small models, $0.20-$0.40 for 32B-72B models and $0.30-$4.00 for 120B-480B models, with images at a $0.01 base rate per 1024x1024 image at 25 steps and audio at $5.00 per 1M characters.
// pros
- A documented free entry tier: the Standard Limits table lists Basic at 60 requests/minute with the requirement "Free account" and a $0 minimum deposit.
- Multiple throughput steps with published thresholds: Pro is 600 requests/minute at a $5+ deposit, Enterprise unlimited, and every tier carries a per-IP cap of 600 requests/minute.
- Transparent metered pricing: small models from $0.10 per 1M tokens, medium 32B-72B at $0.20-$0.40, large 120B-480B at $0.30-$4.00, images from a $0.01 base rate, audio $5.00 per 1M characters.
- Compute credits are sold in $5 minimum increments, hold a 1:1 dollar value and are stated to never expire.
- Inference docs state a Zero Data Retention policy (prompts and responses never stored), TLS 1.3 for API calls and a 99.9% uptime SLA, with SOC 2 Type II described as in progress.
// cons
- The free Basic tier is a rate limit only: there is no published credit grant, and pay-as-you-go usage still draws from a funded balance.
- Official pages conflict on whether free-plan users can call the API at all: the quickstart says "Before launching GPU instances or using Inference, add funds to your account" with a $5 minimum, while the billing FAQ says "inference services are available to free plans".
- The inference catalog is being pruned and is dated: the deprecation list includes Qwen3-Next 80B, GPT-OSS 20B/120B, FLUX.1-dev, SDXL 1.0, Stable Diffusion 1.5/2 and Pixtral 12B, and some named models are old generations (Llama 3.1, Qwen 2.5, DeepSeek V2.5).
- Running out of balance is destructive: GPU instances may be terminated and on-machine data cannot be recovered, with only a 30-day grace period for attached storage volumes.
- Account balance must cover at least one hour of runtime across all instances before a new on-demand instance can be created.
Hyperbolic fits teams that want an OpenAI-compatible endpoint for open-weight models at metered rates and are willing to keep a funded balance, using Basic at 60 requests/minute for light testing before deciding on a $5 deposit. It is best paired with a fallback provider for any workload that cannot tolerate a 429 or a balance-driven instance termination. For pure GPU rental it competes on self-serve access rather than on a free tier, since inference capacity is billed, not granted.
imported analysis — independent third party, not Uprouter editorial · verify before relying on it
Free tier present — value not quantified
This provider offers a free tier, but its terms don’t convert cleanly into a USD figure (e.g. request-count limits or capacity-dependent pools). We refuse to print a misleading zero here. See the note below and the official pricing page for the exact limits.
note: No free tier stated on official pages (checked 2026-09-16): Hyperbolic documents a free "Basic" tier as 60 requests/minute with a Free account, but no dollar credit or trial allowance is published on the pricing or billing pages.
| Plan | Type | Monthly | Input /1M | Output /1M | Note |
|---|---|---|---|---|---|
| Free tier (source-reported) | free | — | — | — | About $1 trial credit on sign-up (phone verification required); occasional promotional credits (e.g. $15 or $25 for hackathons/events); pay-as-you-go afterward, from about $0.0001 per 1K tokens. No long-term fully-free flagship-model quota. |
| Pay-as-you-go | payg | — | $0.200 | $0.200 | Rate for Llama 3.1 405B; other models vary. |
Pricing normalized from public sources — always verify with the provider.
// model prices
| Model | Input /1M | Output /1M | Context |
|---|---|---|---|
| DeepSeek V3 | — | — | — |
| Llama 3.1 405B (BF16, currently the only public 405B Base) | — | — | — |
| Qwen family | — | — | — |
| FLUX / Stable Diffusion (image) | — | — | — |
| Melo TTS (audio) | — | — | — |
$ Does Hyperbolic have a free tier? #
Yes — Hyperbolic offers a free tier. Always confirm current limits on Hyperbolic's official pricing page — free tiers change without notice.
$ How much does Hyperbolic cost per 1M tokens? #
Hyperbolic does not publish simple per-1M-token pricing (it may bill per request, per output, or via a subscription plan). Check their official pricing page for exact figures.
$ Is Hyperbolic safe to use? #
Uprouter rates Hyperbolic as low risk. Risk is moderate. Hyperbolic Labs operates a commercial, publicly branded platform with named executives, an active blog through August 2026, referenced partnerships (Hugging Face, NVIDIA Cloud Provider ecosystem) and a stated compliance roadmap, which makes abandonment risk lower than for a hobby reseller. The free tier is only a rate limit and confers no credit, so there is no promotional balance that can be withdrawn or expire, and compute credits are stated to be non-expiring and 1:1 with dollars. The real exposures are financial accountability and data handling: the service depends on a supplier network of third-party GPU providers, balances reaching zero can permanently destroy on-instance data, and the zero-retention and SOC 2 claims are the vendor's own statements with certification still described as "pursuing". Payment requires a funded balance held with the operator, so unused balance is an unsecured claim if the service is discontinued. Risk ratings are editorial, evidence-based, and never influenced by affiliate relationships.
$ Can I use Hyperbolic through Uprouter Connect? #
Yes — Hyperbolic is Connect-compatible. Add your Hyperbolic API key in Uprouter Connect (encrypted at rest) and route requests to it through one OpenAI- and Claude-compatible API, optionally behind failover aliases. Uprouter meters the traffic in compute at the provider's listed rate.
Raw community sentiment, shown unblended. Votes never feed into objective data fields like price, risk, or status — and never into sort order.
No published reviews yet — be the first.
// more_informationprice history · live status · risk & confidence · code snippets #
price_history
No recorded changes yet.
live_status
best-effort probes · not a guarantee · last 20h ago
| Checked | Status | Latency |
|---|---|---|
| 20h ago | up | 81ms |
| 21h ago | up | 100ms |
| 1d ago | up | 110ms |
| 2d ago | up | 99ms |
| 3d ago | up | 65ms |
| 3d ago | up | 121ms |
| 4d ago | up | 57ms |
| 5d ago | up | 68ms |
| 6d ago | up | 173ms |
| 6d ago | up | 117ms |
| 7d ago | up | 144ms |
| 8d ago | up | 149ms |
| 9d ago | up | 85ms |
| 10d ago | up | 72ms |
| 19d ago | up | 1783ms |
Uprouter runs its own probes; best-effort, not a guarantee. Probe results depend on our network location and cadence — for production, always use the provider’s own status page.
risk_&_confidence
Risk is moderate. Hyperbolic Labs operates a commercial, publicly branded platform with named executives, an active blog through August 2026, referenced partnerships (Hugging Face, NVIDIA Cloud Provider ecosystem) and a stated compliance roadmap, which makes abandonment risk lower than for a hobby reseller. The free tier is only a rate limit and confers no credit, so there is no promotional balance that can be withdrawn or expire, and compute credits are stated to be non-expiring and 1:1 with dollars. The real exposures are financial accountability and data handling: the service depends on a supplier network of third-party GPU providers, balances reaching zero can permanently destroy on-instance data, and the zero-retention and SOC 2 claims are the vendor's own statements with certification still described as "pursuing". Payment requires a funded balance held with the operator, so unused balance is an unsecured claim if the service is discontinued.
Confidence reflects how much verifiable evidence (official docs, probe history, community reports) backs this entry. Risk is our editorial assessment of operator reliability and terms — not financial advice.
code_snippets
curl https://www.uprouter.online/api/connect/v1/chat/completions \ -H "Authorization: Bearer upr_live_YOUR_KEY" \ -H "Content-Type: application/json" \ -d '{
"model": "deepseek-v3",
"messages": [{ "role": "user", "content": "Hello via Hyperbolic" }]
}'
Metered in Uprouter compute — the usage log shows which upstream served each request and its exact cost.
NLP Cloudfree $15.00 est.pricing n/aconnectLow risk
Nomicfree $0.10 est.pricing n/aconnectLow risk
Lightning AIfree $15.00 est.pricing n/aconnectLow risk
AI21 Labsfree $10.00pricing n/aconnectLow risk
Ranked by neutral similarity signals only — live status, Connect compatibility, free-tier shape and free-credit proximity. Commercial relationships never influence this list.