How is this different from renting GPUs on Vast.ai or RunPod?

On Vast.ai or RunPod you rent raw GPUs and run the vLLM stack yourself - CUDA, out-of-memory, cold starts, interrupt recovery, and the serving layer. We charge for the managed OpenAI-compatible endpoint, the SLA, and not babysitting a box. For a team that wants to ship, the flat fee is usually the cheaper total cost.

Can it run Claude, GPT, or Gemini?

No. Those are closed models. The rig runs open-weight models that fit in 128GB - Qwen3.6, DeepSeek V4 Flash, GLM 5.2, Llama, and your own fine-tunes. For the small share of work that needs a frontier closed model, keep a low-limit OpenAI or Anthropic key and we route to it as a fallback.

Is the data actually private?

Yes. Single-tenant hardware, no shared GPU pool, no prompt or completion logging on our side, your API keys, your access logs, and your data is wiped at contract end. Your data does not go to a public API.

When does the flat fee beat paying per token?

Roughly once your team spends more than a few hundred dollars a month on API tokens. Use the token counter to size your real volume, then the budget tool to compare against owning or renting cloud. Below that volume, a per-token API is cheaper and we will say so.

Managed private AI

Managed private AI: skip the vLLM ops

You can run open-weight models yourself on rented GPUs. Many of our readers do. If you would rather not babysit a box, we will run the rig and hand you an OpenAI-compatible endpoint. Flat monthly fee, no per-token bills, and your data stays on dedicated hardware you control.

Inquire to rent a rig See the rig specs

The honest tradeoff

Renting GPUs on Vast.ai or RunPod and running vLLM yourself is the cheapest path per token. It is also a second job: CUDA mismatches, out-of-memory restarts, cold starts, interrupt recovery, and the serving layer on top. A managed rig trades a little cost per token for not doing that.

DIY GPU rental

Cheapest per token. You run vLLM, the queue, healthchecks, and recovery. Best for teams with an AI infra person who likes the work.

Managed rig (us)

Flat fee, OpenAI-compatible endpoint, SLA, and a fallback route to a low-limit frontier key. Best for teams that want to ship, not operate.

When does flat beat per-token?

The break-even lands once your team spends more than a few hundred dollars a month on API tokens. Below that, a per-token API is genuinely cheaper and we will say so. Size your real volume first, then decide.

Step 1

Count your tokens

Paste real prompts and code to see your actual token volume, then estimate the API cost at today's per-token rates.

Step 2

Compare buy vs rent

Set your monthly budget and use case to see what you could own outright versus rent from the cloud, with payback period and tok/s.

What you get

A 128GB rig running open-weight models: Qwen3.6, DeepSeek V4 Flash, GLM 5.2, and your own fine-tunes.
An OpenAI-compatible endpoint. Point the OpenAI SDK, Cursor, Cline, Continue, or LangChain at it and you are running.
Single-tenant hardware hosted in Kingston, Ontario, no shared GPU pool, no prompt or completion logging, data wiped at contract end.
A 99.5% uptime target with healthchecks, auto-restart, and a fallback route to a low-limit frontier key.
No per-token fees, no rate limits, no overage invoices. Month-to-month after 60 days.

Want a rig?

Tell us what you run and your token volume. We will size the right tier and respond within one business day. The full tier comparison and business framing is on the RoboticForce page.