Lium
✓ Maker verified
On-demand GPU pods your agent can start and stop
GPU pod marketplace (RTX 4090/5090 up to H100, B200 and B300) with per-second billing, a CLI, Python SDK and API. Pitched as agent-first: an agent can run `lium up --gpu H100 --ttl 1h`, do the job, and the pod auto-removes within a budget you set. Besides batch jobs, a pod can serve a model behind an OpenAI-compatible endpoint (for example with vLLM). The lium CLI and Python SDK are open source under MIT; the marketplace itself is not. Prices are set per listing by GPU providers and change with availability.
Best for: Agents or developers that need short-lived GPUs for fine-tuning, batch jobs or serving a model behind an OpenAI-compatible endpoint, not for hosting a web app
Visit Lium ↗- Pricing
- Pay-as-you-go, billed per second; prices are set per listing by GPU providers (live at lium.io/pricing; on Oct 24, 2026 listings included H100 80GB from $1.70/hr and 1x B200 at $5.50/hr, cheapest GPUs from $0.30/hr)
- Free tier
- No free tier found; you pay per second for pods
- Open source
- No
- Last verified
- 2026-10-24
Lium pricing breakdown, history and cheaper alternatives →
Sources (checked 2026-10-24): lium.io · lium.io/pricing · github.com/Datura-ai/lium · lium.io/blog/serve-llm-with-vllm-on-lium
Score breakdown
- Fit for the use case12 / 25
- Vibe-coder friendliness10 / 15
- Output quality & reliability9 / 15
- Pricing & free tier11 / 15
- Ecosystem & integrations6 / 10
- Momentum6 / 10
- Code ownership & lock-in6 / 10
Why this score: Cheap, scriptable GPUs with TTL auto-removal are genuinely agent-friendly. It is GPU compute rather than a place to deploy and serve an app, marketplace hardware means reliability varies by provider (listings show ~97% reliability), and it is a smaller platform, so fit and quality score lower in this category.
Based on desk research of official pricing pages and docs; see methodology.
FAQ
- Is Lium free?
- Lium: No free tier found; you pay per second for pods. Paid pricing: Pay-as-you-go, billed per second; prices are set per listing by GPU providers (live at lium.io/pricing; on Oct 24, 2026 listings included H100 80GB from $1.70/hr and 1x B200 at $5.50/hr, cheapest GPUs from $0.30/hr).
- What is Lium pricing in 2026?
- Lium pricing (as of our last verification on 2026-10-24): Pay-as-you-go, billed per second; prices are set per listing by GPU providers (live at lium.io/pricing; on Oct 24, 2026 listings included H100 80GB from $1.70/hr and 1x B200 at $5.50/hr, cheapest GPUs from $0.30/hr). Always confirm on the official site — plans change.
- Is Lium open source?
- No. Lium is not open source.
- What is Lium's PromptToProd score?
- Lium scores 60/100 on PromptToProd (fit, vibe-coder friendliness, quality, pricing, ecosystem, momentum, ownership). See the score breakdown on this page and our methodology.
- What is Lium best for?
- Lium is best for: Agents or developers that need short-lived GPUs for fine-tuning, batch jobs or serving a model behind an OpenAI-compatible endpoint, not for hosting a web app.
- When was Lium last verified on PromptToProd?
- Pricing and details for Lium were last verified on 2026-10-24 from official docs and pricing pages.
Alternatives
Compare all Lium alternatives side by side →
Maker of Lium? Show your score
Free embeddable badge. It updates when the score changes. Adding it is optional and never affects the score. About badges
<a href="https://prompttoprod.app/tool/lium/?ref=badge"><img src="https://prompttoprod.app/badge/lium.svg" alt="Lium on PromptToProd: 60/100" height="20"></a>[](https://prompttoprod.app/tool/lium/?ref=badge)For the Lium team
Claim this listing · Get featured, $49
30 days of a clearly labeled “Featured” card at the top of Hosting & deploy, plus a featured badge. One-time, no auto-renewal. Your score of 60 stays exactly as earned: rankings are never pay-to-rank.