Gemini 3.1 Pro (Preview) is positioned as Google's paid preview tier for long-context reasoning, multimodal input, and tool-heavy developer workflows, with the active model identifier gemini-3.1-pro-preview replacing the now-stored older preview path. A separate gemini-3.1-pro-preview-customtools endpoint is offered specifically for agentic workloads that need to mix bash execution with registered tool calls, making it useful for orchestrating complex pipelines rather than a universal upgrade over the default path. Programmatic access is billed only, so AI Studio serves as the fastest way to validate prompts before wiring billing into production traffic.

The pricing curve is shaped around a 200k-token prompt boundary, charging a baseline input and output rate below that threshold and a higher tier above it, with Batch runs priced at half those rates so large evaluation sweeps become meaningfully cheaper. According to the same sources, the model is reported to more than double its predecessor's reasoning score on ARC-AGI-2, signaling a deliberate step up in analytical depth rather than a cosmetic refresh. That combination of long-context handling, multimodal ingestion, explicit tool orchestration, and tiered pricing makes the preview a practical fit for teams building agent backends, document-and-media analyzers, and research assistants where reasoning quality across very large inputs matters more than minimum cost.