GPT-5.6 Terra sits in the middle of OpenAI's 5.6 generation as a reasoning-focused model designed to balance capability and operating cost, and it is exposed by Vivgrid through its unified model catalog alongside related coding-oriented variants. In Artificial Analysis pre-release evaluation, Terra reached a score of 55 on the Artificial Analysis Intelligence Index at its maximum reasoning effort, positioning it between the higher-end Sol variant and the lighter Luna variant, and it is reported to deliver roughly fifty percent lower Cost per Task than Sol on the same benchmark. That combination of mid-range intelligence score and reduced inference cost frames Terra as a deliberate Pareto trade-off rather than a pure flagship, targeting workloads where strong reasoning is needed but premium-tier spending is not justified.

Because Terra supports text, image, and PDF inputs and emits structured text, it fits naturally into multimodal agent pipelines that ingest mixed documents and return parseable results, with reasoning, tool calling, and structured output enabling controller-style orchestration. The large input budget of the cataloged API limit tokens and 128,000-token output ceiling, against a 1,050,000-token combined context window, make it well suited to long-codebase or long-document tasks such as repository-level analysis, multi-file refactoring, and extended report synthesis. Teams already using Vivgrid for coding agents can drop Terra in alongside Sol and Luna to right-size reasoning effort per task, while developers integrating via the Puter API gain a familiar OpenAI-namespaced endpoint for prototyping the same workloads.