Plans
What Free includes, what each paid plan adds, and which limits change.
Clusy has four plan families: Free, Plus, Pro, and Max. Max comes in 10x and 30x sizes. Prices and the allowance meter in Settings → Billing are the live source of truth; this page explains the capability differences.
At a glance
| Free | Plus | Pro | Max 10x | Max 30x | |
|---|---|---|---|---|---|
| Monthly price | $0 | $12 | $30 | $90 | $200 |
| Models | Auto | + both DeepSeek V4 models | + all open models | + Claude and GPT | Same models as Max 10x |
| GPUs | — | T4 | + L4 and A10 | + L40S and A100 40/80 GB | + H100 and H200 |
| Maximum RAM | 8 GB | 16 GB | 32 GB | 64 GB | 128 GB |
| Reported saved-state allocation | 5 GiB | 30 GiB | 100 GiB | 512 GiB | 2 TiB |
| Live sandboxes | 1 | 3 | 8 | 12 | 24 |
| Included usage | Free allowance | Plus allowance | 3x Plus | 10x Plus | 30x Plus |
GPU entries are cumulative: Pro includes T4, L4, and A10, for example. CPU is fixed at 8 vCPUs and 8 GiB; plan RAM above that is selectable system memory for GPU sandboxes. See Runtimes & sandboxes for VRAM and switching details, and Choosing a model for the exact model list.
The Free plan
Free gives you the Auto model, a CPU sandbox with 8 vCPUs and up to 8 GB of RAM, a 5 GiB reported saved-state allocation, and a monthly allowance.
Two things are worth knowing:
- The built-in model picker is locked to Auto, and GPUs require a paid plan.
- On Free, Auto and CPU consume the allowance. When it is gone, new runs wait for the next monthly reset. On paid plans, Auto, DeepSeek V4 Flash, and CPU form the unmetered floor.
You can still use a frontier model on Free without spending the Clusy model allowance: bring your own key, billed by Anthropic or OpenAI, or use a paired local runner on your own machine. Sandbox and web-tool usage still follows the normal rules.
What each paid plan adds
Plus unlocks named DeepSeek V4 Flash and DeepSeek V4 Pro, the T4 GPU, up to 16 GB of RAM, and three simultaneous sandboxes.
Pro adds every first-party open model—Kimi, GLM, and Qwen—plus L4 and A10 GPUs, up to 32 GB of RAM, and eight simultaneous sandboxes.
Max 10x adds the first-party Claude and GPT models, L40S and A100 GPUs, up to 64 GB of RAM, and ten times the Plus usage allowance.
Max 30x keeps the same model catalog and adds H100 and H200 GPUs, up to 128 GB of RAM, the largest storage volume, and thirty times the Plus allowance.
Every paid plan also gets:
- unmetered Auto, DeepSeek V4 Flash, and CPU runtime,
- and optional pay-as-you-go after the included allowance, with a monthly cap you set.
What “storage” means
The plan volume is for saved kernel state: the variables Clusy checkpoints so a session can resume where it left off. Notebooks, uploads, datasets, workspace files, outputs, and installed packages do not count toward it.
The app reports this allocation for visibility and cleanup. If a new checkpoint is refused, Clusy shows a Storage is full dialog; an existing branch can still replace its current checkpoint. The allocation is separate from pay-as-you-go. In Settings → Storage, you can see which projects hold saved state and clear a project's variables without deleting its notebook or files. See Limits.
What does not change by plan
Sandbox types are available on every plan. The branch cap depends on the runtime rather than the subscription: 3 experiment branches on CPU and 8 on GPU.
Changing plans
Upgrade, downgrade, or cancel from Settings → Billing. Subscriptions and invoices go through Stripe, and the billing portal is linked from the same section.