Salad Cloud
Salad Cloud rents idle consumer GPUs from $0.015 per GPU-hour for retryable AI inference, image generation, transcription, and batch jobs.
Salad publishes the cheapest GPU rates we've seen from a real cloud — $0.015 per GPU-hour at Lowest priority, with per-second billing and free cold starts. The trade is structural: node churn and spot-style interruption are baked into the model, and the good tiers are still consumer RTX cards. If latency is a requirement rather than a nice-to-have, pay more elsewhere.
Verified 17h ago · liveness 78/100 · cite: rightaichoice.com/tools/salad-cloud
- Cost-sensitive teams running retryable, queue-backed inference or batch jobs
- Image generation platforms at scale — Civitai serves 10M images a day on 600+ Salad GPUs
- Speech-to-text and TTS pipelines billed per audio hour or per word
- Kubernetes teams wanting burst GPU capacity via Virtual Kubelet
- Real-time or latency-sensitive serving — Lowest is explicitly the most interruptible tier
- Workloads needing more than 32GB VRAM per card; the published card list tops out at RTX 5090 32 GB
- Buyers who need an SLA, compliance certifications, or guaranteed reserved capacity without a Dedicated contract
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 Salad Cloud if your service has a latency SLO that a preempted container would break, or if procurement requires an SLA and compliance certifications before you can run a single job.
Running above Lowest priority costs more per GPU-hour — High is the premium tier, and picking it to avoid preemption is the single biggest line-item change you can make.
At $0.015 per GPU-hour at Lowest priority with per-second billing, SaladCloud undercuts on-demand AWS GPU pricing by a wide margin and sits at or below Vast.ai and RunPod on published GPU-hour rate. The trade is reliability, not dollars: if you need preemption protection, paying for High priority or a Dedicated contract narrows the gap against those peers. Hyperscaler Spot is the cheaper comparison only when you already have committed cloud spend to burn.
In short
Salad Cloud — Salad Cloud rents idle consumer GPUs from $0.015 per GPU-hour for retryable AI inference, image generation, transcription, and batch jobs. Best for Cost-sensitive teams running retryable, queue-backed inference or batch jobs, Image generation platforms at scale — Civitai serves 10M images a day on 600+ Salad GPUs, Speech-to-text and TTS pipelines billed per audio hour or per word. Plans from $0.005.
No head-to-head comparison for Salad Cloud yet — the alternatives below are the closest matches we list.
What's new in Salad Cloud
Checked 8 days agoAcross the latest 4 updates: 4 changelog entries.
Salad app 1.9.27 adds container image download progress and in-app support tickets
Salad app 1.9.27 adds live download progress for container workload images and a Get Help window for in-app support tickets. AMD GPU workload failures fixed; Star Chef removed.
Salad app 1.9.26 patches critical container workload security vulnerability
Salad app 1.9.26 fixes a critical security vulnerability in the container workload platform per a published advisory. Also corrects incorrect vCPU restrictions on GPU workloads.
Salad app 1.9.24 fixes crash loop in container workloads
Salad app 1.9.24 fixes a crash that caused some container workloads to repeatedly fail and restart, interrupting chopping.
Salad app 1.9.23 adds container workload stop recovery and crypto mining control
Salad app 1.9.23 adds recovery for container workloads stuck in Stopping, fixes cached image state and misattributed container events, and adds an onboarding crypto mining control off by default.
What people actually say about Salad Cloud — is it worth it?
We scanned public community sources for Salad Cloud on Oct 10, 2026 and could not establish that the discussion we found is about this tool rather than something else sharing its name. Only 0 of the posts we fetched could be positively tied to Salad Cloud. 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 Salad Cloud? 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
- Rent GPU instances from $0.015 per GPU-hour at Lowest priority, billed per second while running
- Four priority tiers — High, Medium, Low, Lowest — trade cost against preemption risk
- High-priority groups are never preempted by other workloads
- Medium and Low tiers carry roughly a 30-minute minimum run before reallocation
- vCPU and RAM included with every GPU instance at no extra charge
- Salad Container Engine schedules Docker/OCI containers across distributed nodes it has never seen before
- Automatic workload reallocation when a node disconnects, with 10,000+ replicas and failover
- Hypervisor isolation and a proprietary trust rating that scores node performance and availability
- Published per-class rate card, from RTX 5090 (32 GB) at $0.250/hr to RTX 3060-class at $0.015/hr, Lowest priority
- AI Gateway: OpenAI-compatible LLM endpoint for open-weight models, per-token from $0.06 per million input tokens or a
- Salad AI Gateway serves Qwen at a 262k context window, with data never used for training
- Transcription API at $0.16 per audio hour, 95.1% CommonVoice accuracy
- Text-to-speech up to 6 million words per dollar with OpenVoice on RTX-class GPUs
- Gateway Service: dedicated residential IP tunnels for streaming video
- Virtual Kubelet schedules Kubernetes pods onto SaladCloud nodes
About Salad Cloud
Salad Cloud turns hardware that already exists into a rentable GPU cloud. Rather than financing new data-center racks, Salad orchestrates idle consumer GPUs — RTX 4090s, 5090s, and similar cards sitting in homes, workstations, and small businesses — into an on-demand pool. The company reports 16,000+ daily active GPUs across 190+ countries and 450,000+ cards contributed since 2018, with pricing published openly on its pricing page and billed per second only while an instance is in the running state. Allocation, image download, and cold start are free. One orchestration layer spans three pools. Community Cloud is the shared on-demand network, with a published rate card that runs from $0.250 per GPU-hour for an RTX 5090 down to $0.015 at Lowest priority for additional classes like the RTX 3060 and 3070. Salad Dedicated is reserved bare metal and VMs on contract, led by RTX PRO 6000 Blackwell. Private Cloud runs the same orchestration inside your own organization, pooling idle GPUs you already own into an OpenAI-compatible token service. The scheduling software is the Salad Container Engine, which places Docker or OCI containers onto nodes it has never seen before, verifies and hypervisor-isolates them, streams logs, and reallocates work the moment a node disappears — the vendor cites 10,000+ replicas with failover. Four priority tiers (High, Medium, Low, Lowest) let you trade cost against preemption risk: High is never preempted by other workloads, Medium and Low carry roughly a 30-minute minimum run, and Lowest is the cheap, interruptible batch tier. vCPU and RAM ship with every GPU at no extra charge; CPU-only groups bill at $0.005 per vCPU-hour plus $0.001 per GB RAM-hour. Managed services sit alongside the containers: an OpenAI-compatible AI Gateway for open-weight models from $0.06 per million input tokens (per-token or flat monthly), a Transcription API at $0.16 per audio hour, text-to-speech with OpenVoice, Gateway Service residential IP tunnels, and
Behind the Verdict
Salad's bet is straightforward: most AI workloads don't need to be fast, they need to be cheap. Image generation, transcription backlogs, LoRA training runs, tagging pipelines, GROMACS simulation — the vendor's own benchmarks put 5,243 Flux.1-Schnell images per dollar and 1,000,000 audio hours transcribed for $1,260. On a published rate card that starts at $0.015 per GPU-hour, that math is hard to argue with. We'd reach for this when the queue absorbs the variance. Batch inference behind a job queue, stateless replicas with a restart policy, or anything where re-running a lost unit of work costs less than guaranteeing it never fails. Civitai's case is the proof point: 600+ consumer GPUs serving 10 million images a day. That's not a toy workload. When we'd pass: anything user-facing that can't absorb a cold start or an interruption, and anything that needs to be reserved for a specific date without signing a Dedicated contract. Priority tiers are the dial — High is never preempted by other workloads, but node disconnection is still on the table and SCE reallocates. Medium and Low get roughly 30 minutes before reallocation; Lowest is the cheapest and the most interruptible. VRAM is the other ceiling. The published card list tops out at 32GB on RTX 5090, with most classes at 12–16GB. Large models that need more per card are out unless you shard. The closest alternatives are Vast.ai and RunPod — similar per-hour economics, more conventional data-center supply, and generally tighter latency control. Salad's edge is scale and price floor; theirs is predictability. AWS EC2 Spot or a dedicated cloud wins if you need an SLA, compliance paperwork, or certification coverage. A practical caveat: the network is consumer hardware, so engineering hygiene matters more than usual.
Researching Salad Cloud? 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 Salad Cloud actually fits — and what changes day-one when you adopt it.
Push a Docker image to the Salad Container Engine through the API, set the container group to Lowest priority with 40 replicas, and let SCE schedule them across RTX 4090 nodes worldwide.
Outcome: Billing starts only when each instance reaches the running state, and the platform reallocates any replica whose node drops — the queue absorbs the interruption instead of failing the request.
Send audio files to the Transcription API at $0.16 per audio hour instead of standing up GPU instances, and compare weekly volume against a self-hosted Whisper deployment.
Outcome: Published 95.1% CommonVoice accuracy lets you judge quality before committing, and pay-as-you-go with no minimum means the comparison costs one batch.
Install Virtual Kubelet and schedule overflow pods onto SaladCloud when your own cluster saturates, keeping the same Kubernetes control plane.
Outcome: Burst capacity arrives without a new cloud account or a rewrite, and pods that get reallocated simply reschedule onto the distributed pool.
Use Cases
- Serve image generation inference at scale, reducing costs up to 90% — Civitai runs 10M images/day on 600+ Salad GPUs
- Run zero-knowledge proof computations with 77% savings versus data centers
- Power voice AI applications with distributed GPU nodes
- Perform batch processing across thousands of low-cost GPUs
- Collect data via residential IP tunnels through Salad Gateway Service
- Scale molecular dynamics simulations like GROMACS on thousands of low-cost GPUs
- Transcribe audio at $0.16 per audio hour via the Transcription API
- Run OpenClaw agents on distributed GPU capacity
Models Under the Hood
as of 2026-10-16
Limitations
- SaladCloud rents idle consumer-grade GPUs, so performance and availability vary by node, and higher priority tiers can take over lower-priority workloads.
- Unreserved capacity is subject to interruption and automatic reallocation.
- Capacity is exposed through an OpenAI-compatible LLM endpoint and a transcription API, with the private AI Cloud and Salad Dedicated offerings positioned for reserved or enterprise use.
- Pricing is billed per second, and Salad quotes up to 90% less than hyperscaler pricing, but no SLA or compliance certification is stated in the provided evidence.
as of 2026-09-29
Verification history
We have re-verified Salad Cloud 19 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-checked, vendor evidence unchanged
- — 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-checked, vendor evidence unchanged
- — 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 19 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 Salad Cloud tier: who it actually fits, and what it adds vs. the previous tier. Cross-reference the cost calculator above for projected annual outlay.
GPU Instances — Community Cloud (per GPU-hour by priority)
from $0.015/hr
CPU Instances — Community Cloud
$0.005/vCPU-hour + $0.001/GB RAM-hour
AI Gateway
from $0.06 per million input tokens
Transcription API
$0.16 per audio hour
Volume Discounts
Custom
Salad Dedicated
Custom
Ideal for
Buyers who need reserved capacity and RTX PRO 6000 Blackwell hardware instead of interruptible community nodes.
What this tier adds
Reserved bare metal and VMs on contract, quoted per allocation and term, with same API and portal as Community Cloud.
Private Cloud
Custom
Ideal for
Organizations with a fleet of workstations and servers sitting idle that want the orchestration on their own infrastructure.
What this tier adds
Licensed per deployment, runs Salad's orchestration inside your organization and turns your own hardware into an OpenAI-compatible token service.
Where the pricing makes sense
The company stage and team size where Salad Cloud's pricing actually pencils out — and where peers do it cheaper.
At $0.015 per GPU-hour at Lowest priority with per-second billing, SaladCloud undercuts on-demand AWS GPU pricing by a wide margin and sits at or below Vast.ai and RunPod on published GPU-hour rate. The trade is reliability, not dollars: if you need preemption protection, paying for High priority or a Dedicated contract narrows the gap against those peers. Hyperscaler Spot is the cheaper comparison only when you already have committed cloud spend to burn.
Setup time & first value
How long it actually takes to get something useful out of Salad Cloud — broken out by persona, not the marketing-page minute.
Container work is the fast path: with a Docker image already built, an API key, and a container group definition, a first deployment reaching the running state typically takes under an hour, and you are only billed once instances run. Kubernetes teams using Virtual Kubelet add a node to an existing cluster in roughly the same window. Private Cloud on your own hardware is a deployment project
Switching to or from Salad Cloud
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From AWS EC2 Spot: containerize the workload, set Lowest or Low priority to mirror Spot's interruptibility, and keep a queue in front of it so preemption is a retry rather than an outage.
- →From RunPod: reuse the same Docker image and switch the schedule call to the Salad Container Engine API, mapping your old instance type to an equivalent RTX class.
- →From local A100s: Blend's reported 3X scale at half the cost is the template — move batch and image-generation jobs to Community Cloud first and keep latency-critical paths local.
- →From self-hosted Whisper: route audio through the Transcription API at $0.16 per audio hour and retire the GPU sitting idle between batches.
- →From Vast.ai: keep the retry semantics you already built and change only the provider API — the interruption model is comparable, the published rate is lower.
- ↗To AWS EC2 or a hyperscaler: the path back is for workloads that outgrew the interruption model or need an SLA and certifications, and it is mostly a re-containerization plus a queue rewrite.
- ↗To RunPod or Vast.ai: if you want tighter control over which GPU you land on and comparable per-hour economics, the container images move unchanged.
- ↗To Salad Dedicated: the in-family upgrade when Community Cloud's interruption model no longer fits — same API and portal, reserved RTX PRO 6000 Blackwell on contract.
- ↗To Private Cloud: for organizations that want the same orchestration running on hardware they already own, licensed per deployment.
Integrations
Resources & Guides
- Resourcesalad.com
Home - Salad Chef Community
Salad Chef Community The latest news on the Salad app, the latest rewards and everything that’s cooking in the kitchen. Exclusive Deals on Discord Bots & Web Hosting by Cybrancee Online presence has never been more important, and managing web hosting for your various person
- Resourcesalad.com
Home - Salad Chef Community
Salad Chef Community The latest news on the Salad app, the latest rewards and everything that’s cooking in the kitchen. Exclusive Deals on Discord Bots & Web Hosting by Cybrancee Online presence has never been more important, and managing web hosting for your various person
Tutorials & Learning
YouTube returned 6 videos for “Salad Cloud”, and we withheld 6: 6 could not be judged, because “Salad Cloud” 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 Salad Cloud.
Official links
Tools that pair well with Salad Cloud
Common stack mates teams adopt alongside Salad Cloud, with the specific reason each pairing earns its keep.
Inference Engine by GMI Cloud
GMI Cloud Inference Engine runs text, image, video, and audio models behind one OpenAI-compatible API on dedicated NVIDIA GPUs.
MimicPC
Browser-based cloud that runs 20+ pre-installed open-source AI apps for image, video, and audio generation
Stable Horde
Free AI image and text generation on Stable Horde — volunteer GPUs, no account needed.
Alternatives to Salad Cloud
View all
Inference Engine by GMI Cloud
GMI Cloud Inference Engine runs text, image, video, and audio models behind one OpenAI-compatible API on dedicated NVIDIA GPUs.
MimicPC
Browser-based cloud that runs 20+ pre-installed open-source AI apps for image, video, and audio generation
Stable Horde
Free AI image and text generation on Stable Horde — volunteer GPUs, no account needed.
Frequently Asked Questions
Categories
Best-of guides
Topics
Used Salad Cloud? Help shape our editorial sentiment research.