# Marketplace

C3 aggregates GPU capacity from multiple data centers. When you submit a job, we find available compute at competitive rates—no need to manage cloud accounts or hunt for capacity yourself.

## Available GPUs[​](#available-gpus "Direct link to Available GPUs")

| GPU class  | VRAM                         | Best for                                                                         |
| ---------- | ---------------------------- | -------------------------------------------------------------------------------- |
| L40 class  | Current range from `c3 list` | Recommended for new jobs, training, inference, and general-purpose GPU workloads |
| A100 class | Current range from `c3 list` | Larger memory and higher-throughput CUDA workloads                               |
| H100 class | Current range from `c3 list` | Highest-end accelerator workloads when H100 capacity is enabled                  |

Omit `gpu` in your `.c3` config, or set `gpu: l40`, `gpu: a100`, or `gpu: h100`, to let C3 choose within that class. The job records the concrete profile it lands on, such as `l40s-d-32x192` or `h100-80gb`, and bills the selected provider/profile rate. Run `c3 list` for the public GPU-class view with current pricing/vRAM ranges and C3-level availability. Run `c3 list -al` to see the concrete profile codes inside each class. Provider/region offerings are not exposed before a job is matched. A4000 is retired from C3 public stock.

## Pricing[​](#pricing "Direct link to Pricing")

### Compute[​](#compute "Direct link to Compute")

You're billed per second of actual compute time—not for time spent queuing or provisioning. New accounts receive £10 in free credits. Check your balance:

```
c3 balance
```

### Plans[​](#plans "Direct link to Plans")

Subscriptions control maximum GPU concurrency. Compute is still billed per second, and storage usage is tracked separately for usage-based billing. Every new account starts with £10 in compute credits.

| Plan | Price      | Seats     | GPU max concurrency |
| ---- | ---------- | --------- | ------------------- |
| Free | £0/month   | 1         | 1                   |
| Pro  | £45/month  | 1         | 10                  |
| Team | £185/month | Unlimited | 50                  |

Team self-service is coming soon. C3 can grant Team access to selected accounts before public checkout is enabled.

Manage your plan anytime:

```
c3 upgrade              # Interactive picker
c3 upgrade pro          # Subscribe to or switch to Pro
c3 upgrade team         # Team is visible but coming soon
c3 upgrade free         # Cancel subscription (takes effect at period end)
c3 upgrade free --yes   # Non-interactive cancel (for scripts)
c3 upgrade free -y      # Short form of --yes
```

Free to Pro opens Stripe checkout. Downgrades and cancellations take effect at the **end of your current billing period**. To reverse a pending downgrade or cancellation before it takes effect, run `c3 upgrade <current-tier>`.

## How jobs run[​](#how-jobs-run "Direct link to How jobs run")

1. **PENDING** — Job submitted, waiting for a GPU
2. **SCHEDULING** — Assigning to a machine
3. **RUNNING** — Your script is executing
4. **COMPLETED** — Done. Download results with `c3 pull`
5. **SYNCED** — Results downloaded locally

Failed or cancelled jobs show as **FAILED**, **CANCELED**, or **TIMED\_OUT**.

## Startup behavior[​](#startup-behavior "Direct link to Startup behavior")

C3 routes each job to available capacity for the requested GPU profile. Startup time depends on current provider availability, requested hardware, data transfer, and environment setup.

## Providers[​](#providers "Direct link to Providers")

By default, `c3 deploy` asks the C3 marketplace to choose a route before job submission. The control plane then makes the authoritative selection from enabled provider/GPU offerings using stock, locality, price, and provider spend guard state.

Use `provider:` in `.c3`, `c3 deploy -p <provider>`, or `C3_PROVIDER` only when you need to pin a job to a specific provider such as `nextgen` or `nebius`. Provider availability is still gated by operator configuration, live inventory cleanup, spend controls, and stock, so a pinned provider can be rejected while auto-routing succeeds on another provider.

Provider pins are an advanced workflow for provider ids documented by C3 support or runbooks. Public `c3 list` and `/v2/hardware` responses do not expose provider/region offerings, stock, warm-pool capacity, cache affinity, or provider-specific prices before a job is matched. After submission, `c3 deploy` prints the matched provider, region, and concrete GPU profile from the created job.

Operators can inspect provider readiness with:

```
c3 admin balance <provider>
c3 admin provider-status <provider>
```