Lilac
Lilac rents idle enterprise GPU clusters on a spot market and runs the racks for the data centers that own them.
Lilac's real product isn't the marketplace, it's the operations layer — burn-in, GPU replacement, RMA chasing, and one billing record that survives a lender audit. The $1.00/hr spot H100 and $1.50/hr H200 numbers are aggressive for retry-tolerant batch and agentic loops on open-weight models. Pass if you need contractual latency guarantees, frontier proprietary models, or if your security review forbids third-party workloads on your own racks.
Verified 2d ago · liveness 79/100 · cite: rightaichoice.com/tools/lilac
- Startups and engineering teams running cheap inference on open-weight models
- GPU owners with idle enterprise clusters who want to monetize spare capacity without shipping hardware out
- Data centers that need burn-in, GPU replacement, RMA handling, and auditable usage statements
- Teams running batch GPU jobs without managing their own cluster
- Real-time production services that need contractual guaranteed throughput or latency SLAs
- Teams dependent on proprietary frontier models like GPT, Claude, or Gemini
- Enterprises whose security review forbids third-party workloads on their own infrastructure
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip Lilac if your workload is live customer traffic that needs contractual guaranteed throughput or latency SLAs, or if your stack depends on GPT, Claude, or Gemini rather than open-weight models.
Subscription credits are only worth up to 12x when supply is idle — during busy periods your $10/mo, $30/mo, or $100/mo buys far less compute than the headline multiplier suggests.
Lilac fits seed-to-Series-B teams and independent ML engineers who can schedule around spot availability: $10/mo Basic, $30/mo Pro, and $100/mo Max subscription credits, plus $1.00/hr H100 and $1.50/hr H200 batch and ~$2.00/hr dedicated H100 reservations. That undercuts most mainstream managed GPU clouds on comparable hardware. It is not the cheapest raw-iron path — bare-metal resellers and self-managed clusters can go lower — but you are buying testing, 24/7 monitoring, and one auditable
In short
Lilac — Lilac rents idle enterprise GPU clusters on a spot market and runs the racks for the data centers that own them. Best for Startups and engineering teams running cheap inference on open-weight models, GPU owners with idle enterprise clusters who want to monetize spare capacity without shipping hardware out, Data centers that need burn-in, GPU replacement, RMA handling, and auditable usage statements. Free to start; paid plans from $1.
What's new in Lilac
Checked yesterdayAcross the latest 2 updates: 1 launch and 1 news mention.
GPU quality is something to plan for
Lilac published a post on GPU quality planning, framed as guidance on what buyers should account for in their infrastructure planning.
Introducing the new Lilac
Lilac announced a new version of its platform. No further details were available in the indexed entry.
What people actually say about Lilac — is it worth it?
We scanned public community sources for Lilac on Jul 3, 2026 and could not establish that the discussion we found is about this tool rather than something else sharing its name. Our own analysis of that scan says the posts were off-subject. Rather than publish a sentiment score built on the wrong subject, we publish nothing here and re-run the scan.
Viability Score
How well maintained and how widely used is Lilac? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: October 2026
How we score →Key Features
- Spot market for idle enterprise GPUs across H100, H200, B200, and B300 clusters
- Serverless inference through an OpenAI-compatible API endpoint
- Pay-per-token inference pricing with cache-read discounts on repeated context
- Batch container jobs billed per second at H100 $1.00/hr and H200 $1.50/hr
- Monthly subscription credits worth up to 12x when supply is idle
- Dedicated GPU clusters with 1, 6, or 12+ month reservation terms
- Lilac Flex auto-monetizes idle windows inside a dedicated reservation
- Capacity exchange to relist or transfer eligible GPU commitments
- Lilac Kubernetes operator for GPU owners with 70% revenue share
- Server burn-in before customer handover plus failed-GPU replacement and vendor RMA handling
- Self-serve signup and API keys
- Quantization support for FP8, INT4, and NVFP4 to cut memory cost
- Serves open models Kimi K2.6, GLM 5.1, Gemma 4, and MiniMax M2.7
- Delivery, monitoring, and statement console for cluster operators
- Usage records that back invoices, SLA credits, and lender audits
About Lilac
Lilac is a two-sided GPU infrastructure platform. Data centers and GPU owners keep their hardware in-house, deploy a Lilac Kubernetes operator, and take 70% of what the cluster earns — Lilac burns in every server before a customer touches it, drains and replaces failed GPUs before anyone files a ticket, chases vendor RMAs, and produces the single usage record behind every invoice and SLA credit. Buyers get leased capacity without the operational mess. On the demand side, AI teams lease H100, H200, B200, and B300 capacity either through a spot market or as dedicated reservations with 1, 6, or 12+ month terms. Batch container jobs bill per second at H100 $1.00/hr and H200 $1.50/hr. Serverless inference runs on an OpenAI-compatible endpoint with pay-per-token pricing and cache-read discounts on repeated context. The model catalog is open-weight only — Kimi K2.6, GLM 5.1, Gemma 4, and MiniMax M2.7, which landed through a May 2026 MiniMax partnership. Quantization support covers FP8, INT4, and NVFP4. Subscriptions start at $10/mo Basic, $30/mo Pro, and $100/mo Max, with credits worth up to 12x when supply is idle. A new platform release shipped July 21, 2026, and the Saturn Cloud tie-up was extended July 15, 2026 to add GPU capacity for Token Factory. The positioning is deliberately unglamorous: Lilac sells the boring half of the neocloud business — delivery tracking, burn-in, monitoring, statements, and RMA chasing — rather than only the racks. That puts it against managed GPU clouds on the demand side and against pure marketplace listings on the supply side. Lilac is backed by Y Combinator and partners with Saturn Cloud, Osmosis, and Z.ai.
Behind the Verdict
Most neocloud pitches stop at the racks. Lilac starts there and handles the parts nobody wants to own: burning in servers before handover, draining failed GPUs before a ticket gets filed, and chasing vendor RMAs. If you run a smaller data center trying to turn idle enterprise hardware into revenue, that's the pitch that matters — deploy the Kubernetes operator, keep your racks, take 70% of what the cluster earns. We'd reach for Lilac when the workload tolerates retries. Batch container jobs at $1.00/hr H100 and $1.50/hr H200, plus cache-read discounts when agentic loops re-read the same context, are cheap enough that a redirected job costs less than a premium tier costs outright. Subscription credits worth up to 12x when supply is idle are the same idea — pay for a floor, consume it when the market is quiet. Where it bites: the catalog is open-weight. Kimi K2.6, GLM 5.1, Gemma 4, MiniMax M2.7. If your product is pinned to GPT, Claude, or Gemini, Lilac isn't the answer, and no amount of price advantage changes that. Teams needing contractual throughput or latency SLAs for live production should look at managed GPU clouds with published guarantees instead. The closest alternative is a standard GPU marketplace or a managed cluster provider. Those rent you capacity. Lilac also sells the delivery, monitoring, and statement machinery to the owner of the hardware, which is why the two sides of it feel like different companies. That's a real differentiator, and also a reason to check fit carefully — you're buying into their operating model, not just a rate card. One caveat on the July 21, 2026 platform refresh: the announcement didn't disclose detail, so don't plan migrations around it until you've seen it. In practice, the supplier side is the more interesting product.
Researching Lilac? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Real-world workflow fit
Concrete scenarios for the personas Lilac actually fits — and what changes day-one when you adopt it.
Signs up self-serve, gets an API key, and points an existing OpenAI SDK client at Lilac's endpoint to serve Kimi K2.6 or GLM 5.1 for an internal agent that re-sends the same long system prompt on every turn.
Outcome: Long-context calls hit cached prefixes instead of full pay-per-token rates, and the $30/mo Pro subscription credits stretch up to 12x whenever supply is idle, keeping monthly inference spend below what a mainstream managed endpoint would charge.
Submits containerized batch jobs to the spot market at H100 $1.00/hr or H200 $1.50/hr, with per-second billing and checkpointing between steps so a preemption costs a retry rather than the run.
Outcome: Batch throughput lands at a fraction of on-demand cloud GPU rates, and the per-second meter means idle seconds between job stages are not billed.
Installs the Lilac Kubernetes operator, keeps the hardware in-house, and lets Lilac test the cluster before billing starts and monitor it 24/7 afterward.
Outcome: The cluster earns revenue at a 70% supplier share, and the single Lilac record backs invoices, SLA credits, and lender audits — useful evidence of utilization when financing the next hardware purchase.
Use Cases
- Run inference on open frontier models with pay-per-token pricing, compatible with the OpenAI SDK.
- Submit containerized batch jobs to run on idle GPUs at $1–$1.50 per hour.
- Monetize underutilized GPU clusters by installing the Lilac Kubernetes operator and earning 70% revenue share.
- Reserve dedicated clusters via brokered quotes for longer-term projects with 1, 6, or 12+ month terms.
- Use subscription credits that multiply in value when GPU supply is idle, stretching your budget.
- Deploy Token Factory model serving on Lilac capacity via the Saturn Cloud partnership.
- Cut inference cost for long-context and agentic loops with cache-read pricing.
- Relist or transfer eligible GPU commitments through the capacity exchange when you over-reserve.
Models Under the Hood
as of 2026-10-17
Limitations
- The model catalog appears limited to open-weight models; the live evidence names only Kimi K2.6 (live on Lilac) and MiniMax M2.7 (added via the MiniMax partnership).
- Capacity is spot-based and brokered through partners, so GPU availability and quotes depend on third parties rather than a fixed Lilac inventory.
- The platform targets data centers and enterprises running clusters of thousands of nodes, and public materials are mostly announcements without detailed specs or disclosed terms/capacity figures.
as of 2026-10-09
Verification history
We have re-verified Lilac 10 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
Showing the 6 most recent of 10 verification passes.
Free to cite with attribution — this page re-verifies continuously.
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Plans compared
For each published Lilac tier: who it actually fits, and what it adds vs. the previous tier. Cross-reference the cost calculator above for projected annual outlay.
Basic
$10/mo
Pro
$30/mo
Max
$100/mo
Batch Jobs (usage)
$1.00/hr H100; $1.50/hr H200
Ideal for
Data and ML engineers running containerized jobs that can tolerate spot preemption and retry.
What this tier adds
Moves you off credits to per-second usage billing at H100 $1.00/hr and H200 $1.50/hr across H100, H200, B200, and B300 spot capacity.
Dedicated Clusters (usage)
~$2.00/hr H100
Ideal for
Teams that need dedicated H100 capacity for a defined project window with 1, 6, or 12+ month terms.
What this tier adds
Adds dedicated reservations at roughly $2.00/hr H100, plus Lilac Flex to auto-monetize idle windows and a capacity exchange to relist or transfer eligible commitments.
Supplier Revenue Share
70% of revenue
Ideal for
Data centers and GPU owners holding enterprise clusters that are not fully utilized.
What this tier adds
The supply-side model rather than a buyer plan: keep hardware in-house, run the Kubernetes operator, and keep 70% of revenue with testing before billing plus 24/7 monitoring.
Where the pricing makes sense
The company stage and team size where Lilac's pricing actually pencils out — and where peers do it cheaper.
Lilac fits seed-to-Series-B teams and independent ML engineers who can schedule around spot availability: $10/mo Basic, $30/mo Pro, and $100/mo Max subscription credits, plus $1.00/hr H100 and $1.50/hr H200 batch and ~$2.00/hr dedicated H100 reservations. That undercuts most mainstream managed GPU clouds on comparable hardware. It is not the cheapest raw-iron path — bare-metal resellers and self-managed clusters can go lower — but you are buying testing, 24/7 monitoring, and one auditable
Setup time & first value
How long it actually takes to get something useful out of Lilac — broken out by persona, not the marketing-page minute.
For buyers, the July 2026 platform refresh enables self-serve signup, so API keys land in minutes rather than after a sales conversation. For batch jobs, budget time to containerize the workload and confirm your job tolerates preemption. For suppliers, expect the longer path: install the Lilac Kubernetes operator, then wait through cluster testing before billing starts.
Switching to or from Lilac
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From a mainstream managed GPU cloud: repoint your OpenAI-compatible client at Lilac's endpoint and switch serverless calls to open-weight models (Kimi K2.6, GLM 5.1, Gemma 4, MiniMax M2.7).
- →From self-managed Kubernetes GPU nodes: replace your own scheduling with the Lilac Kubernetes operator, or lease spot capacity instead of owning hardware.
- →From on-demand cloud instances for batch: containerize the job and submit to Lilac's per-second batch billing at H100 $1.00/hr or H200 $1.50/hr.
- →From an idle in-house cluster: deploy the Lilac Kubernetes operator to convert spare capacity into revenue at a 70% share.
- ↗To a mainstream managed GPU cloud: restore the proprietary frontier model endpoints Lilac cannot serve and re-provision on-demand capacity for live production traffic.
- ↗To bare-metal GPU resellers: if you can absorb cluster operations yourself, the raw hourly rate can go below Lilac's brokered quotes.
- ↗To a fully managed training provider: Lilac leases inference and batch capacity, so training-heavy workloads need a platform built for that.
Integrations
Resources & Guides
Tutorials & Learning
YouTube returned 6 videos for “Lilac”, and we withheld 6: 6 could not be judged, because “Lilac” is a single word that other videos use for other things. We are showing none, because we could not prove any of them are about Lilac.
Official links
Featured Head-to-Head Comparisons
Lilac vs Spider Cloud
These two tools solve completely different problems. Spider Cloud is essential for any AI pipeline that needs fresh, structured web data at scale — its Rust engine, AI extraction, and catalog of 1,000+ scrapers make it a no-brainer for RAG and agent workflows. Lilac is a specialized compute marketplace for teams that either have idle GPUs to sell or want the cheapest possible inference on frontier models. Choose based on your data bottleneck: fetching external data (Spider Cloud) vs. running models cheaply (Lilac). They can even complement each other.
Lilac vs Voyage Ai
Voyage AI is the clear choice for enterprises building RAG systems that demand domain-specific accuracy, long-context (32K tokens), and compliance (SOC 2, HIPAA). Lilac suits cost-conscious teams or GPU owners wanting to monetize spare capacity, but it lacks retrieval specialization and enterprise trust. Pick Voyage for search quality; pick Lilac to run cheap inference on idle hardware.
Lilac vs Temporal Ai
Choose Temporal if you need reliable, stateful orchestration for AI agents and microservices where failure recovery is critical. Choose Lilac if your priority is low-cost inference or monetizing idle GPU capacity. They solve fundamentally different problems: workflow durability vs. compute cost optimization. Temporal’s freemium model and open-source SDKs make it accessible; Lilac’s pay-per-token with cache-read pricing suits high-volume inference.
Popular in GPU Cloud & Model Inference
Rain AI
Rain AI is a pre-production hardware company building energy-efficient, brain-inspired analog in-memory chips for ultra-low-power AI inference at the edge.
Recogni
Recogni's Tensordyne Napier is a rack-scale AI inference system running on logarithmic math silicon for multi-trillion-parameter MoE
Spectral Labs SGS-1
Decentralized AI inference with sub-5ms latency and verifiable compute
Frequently Asked Questions
Categories
Topics
Used Lilac? Help shape our editorial sentiment research.