Answer hub
Answers to common questions about running AI workloads
Get concise answers about Jungle Grid, AI inference, compute planning, and moving from an experiment to a repeatable workload.
- Best for
- Fast orientation
- Start here when you want a clear answer before going deeper.
- Primary topics
- Inference + compute
- Covers beginner AI compute questions and Jungle Grid basics.
- Next move
- Open a guide
- Use the answer here to choose the deeper page that matches your actual workload.
Quick answers
Start with the question you need answered
Use these answers to understand the basics, then open the relevant guide when you need more detail.
For a specific workload, check the model pages or pricing. For implementation steps, go directly to the docs.
Jungle Grid
Using Jungle Grid
What is Jungle Grid in plain English?
Jungle Grid is a workload execution platform for AI jobs. Instead of making you choose raw GPU infrastructure by hand for every run, it routes workloads across distributed capacity based on fit, cost, latency, and reliability.
Do I need to pick a GPU manually to use Jungle Grid?
That is not the main workflow. Jungle Grid is built around workload intent first, then routed execution underneath it so developers do not have to manage fragmented GPU choices directly for every job.
Why would someone use Jungle Grid instead of piecing together provider workflows?
Because manual provider workflows do not scale cleanly once workloads, models, or failure cases multiply. Jungle Grid is designed to keep the workload interface stable while routing decisions change underneath it.
AI basics
Inference, training, and getting started
Do beginner developers need their own GPU to start building with AI?
Usually no. Most beginners can start with cloud or routed capacity while they learn the workload and validate whether the project needs dedicated hardware later.
What is AI inference?
AI inference is when a trained model takes new input and produces an output. If your app sends prompts, documents, or requests to a model and gets results back, that is usually inference.
What is the difference between inference and training?
Training creates or updates the model. Inference uses the trained model to answer new requests. Most AI product features start as inference workloads, not training workloads.
Planning and compute
Choosing compute and moving into production
What GPU do I need for my AI app?
The answer depends on the workload. Model size, precision, concurrency, and latency target matter more than brand-first hardware shopping. The right route is the smallest healthy one that fits the app's needs.
When should a team stop experimenting and use a proper execution layer?
Usually when the workload becomes repeatable and operational quality starts to matter: stable latency, cost control, failure recovery, and less manual GPU decision-making.