Models, inference, and hosting.
Short explainers for developers calling the API and operators putting GPUs on the network. Written to be useful, not fluffy.
-
Developers
Drop-in OpenAI SDK → Scalattice
Swap base URL and API key. Copy-paste Python and curl. First completion in under two minutes.
-
Models
Choosing a model from the catalog
When to pick Llama 3.3 70B vs Qwen3 mid-sizes, Gemma, DeepSeek R1, or the coder. VRAM and job shape matter more than brand names.
-
Inference
What inference actually is
Tokens, context, streaming, and why per-million pricing shows up on your bill. A plain-language map of the Scalattice API path.
-
Hosting
Host inference on your GPU
Agent install, disk for weights, schedules, and what “in the pool” means before you earn on catalog jobs.
FAQ
What is Scalattice?
Managed production inference across a distributed GPU network. Developers use an OpenAI-compatible API. Operators host catalog models and earn a majority share of token revenue on completed jobs.
Which models can I use?
See the live catalog in Scalattice Cloud and on pricing. Current lineup includes Llama 3.3 70B, Qwen3 32B / 14B / 8B, Gemma 3 27B, DeepSeek R1 7B, and Qwen2.5 Coder 7B.
Can I earn money hosting inference?
Yes, if your machine fits a catalog model and stays ready in the pool. Start with the hosting guide and providers.
Is the API OpenAI compatible?
Yes. Swap base URL and API key in the OpenAI SDK. Step-by-step: Drop-in OpenAI SDK → Scalattice. Also see developers and Cloud docs.