AI Hosting documentation
Choose a model and hardware configuration, review an estimate, and request a quote for hosted AI workloads running on FusionPointAI infrastructure.
- Define the workloadRecord model, context, concurrency, and serving needs.
- Compare modelsCheck public model releases and their published evidence.
- Check hardwareCompare available GPU configurations against memory needs.
- Review a recommendationSubmit workload requirements for a tailored starting point.
- Request a quoteReview the proposed scope and submit it through the available quote flow.
Prepare workload requirements
Start with the task the model must perform and the data sensitivity of the workload. Note the model family or capabilities you require, expected context length, concurrent requests, input and output patterns, and whether the system needs dedicated serving or supporting customization. Separate firm requirements from preferences. These details help make catalogue comparisons meaningful and give the recommendation flow enough context without assuming that one GPU or model fits every deployment.
- Write down expected context length and concurrent workload.
- Identify data handling, region, and deployment constraints.
- Mark required capabilities separately from optional preferences.
Compare models and hardware
Use the public model and hardware catalogues to narrow the candidate set before requesting a configuration. Review release-level information rather than relying on a family name alone, and compare hardware capacity with the workload you described. A catalogue listing is a selection aid, not a performance promise. Where a workload depends on an unlisted model, special software, or a specific capacity profile, explain that requirement rather than guessing at compatibility.
- Open a model release and review its published details.
- Compare available GPU configurations and memory information.
- Keep a short list of candidates with the reason each may fit.
Get a recommendation and estimate
The recommendation and estimate pages help translate requirements into a candidate service scope. Enter workload details using the fields presented on the page and review the returned assumptions before treating an estimate as useful. An estimate is not a confirmed reservation or a final commercial commitment. If the proposed configuration omits a required model, deployment constraint, or operational need, revise the inputs or describe the gap when requesting a quote.
- Submit the workload details on the recommendation page.
- Review configuration assumptions and any stated limitations.
- Use the estimate flow to compare plausible configurations.
Request and review a quote
When the public flow offers a suitable configuration, request a quote and review its scope before proceeding. Sign in if the quote action requires an account. Confirm the selected model or workload, hardware, recurring and one-time items shown, and any customer responsibilities. Do not infer delivery dates, availability, service levels, or included operations unless they are explicitly stated in the quote or agreed documentation. Ask the sales team to clarify any unresolved dependency before accepting.
- Sign in when prompted and submit the quote request.
- Compare the written quote with your original requirements.
- Ask for clarification on exclusions or assumptions before acceptance.
Frequently asked questions
Does AI Hosting run on my own web server?
No. AI Hosting is for workloads running on FusionPointAI infrastructure. It is not a web-hosting or VPS provisioning product.
Can I choose any model or GPU?
Selection is limited to the models, releases, and hardware options presented by the live catalogues and configuration flow. Ask about requirements not represented there.
Is a recommendation or estimate a reservation?
No. Treat it as planning information. The quote and subsequent written agreement determine the proposed scope and commercial terms.
What information should I include in a quote request?
Describe the model or task, context and concurrency needs, data constraints, desired deployment, and any operational requirements that affect scope.