Run · Inference playground
Prepare a useful request, then inspect the sample record
The SAVRN Cloud playground starts with a research purpose. Choose Evidence review, Structured extraction or Research code, provide public or invented material, and inspect the request before running a sample. A prepared Qwen candidate records your intended model choice; a fictional demonstration fixture produces the browser-local sample. Connected model execution is Coming soon.
Evidence review: stay within supplied material
Use this starter to frame a question around excerpts you supply. Ask for the claims supported by those excerpts, references back to the supplied passages and a clear statement of missing evidence. The preview does not browse a URL or verify an external publication. A future model result would still need a researcher to check its citations and distinguish direct evidence from inference.
Structured extraction: make missing information visible
Define the fields you need and supply a small public or invented text. The starter asks for null when a value is absent rather than filling the gap with a guess. This demonstrates an output requirement, not qualified schema enforcement. In a connected workflow, the result would need field validation, source checks and explicit handling of ambiguous or conflicting values.
Research code: draft with a test plan
Describe a small analysis task, an existing code fragment or a proposed transformation. Ask for assumptions, an explanation and checks to run before relying on the change. Neither supplied code nor generated code is executed here. A browser sample is not evidence that a program is correct, safe or reproducible; testing belongs in a separately approved execution environment.
Read the candidate and the fixture separately
Preparing a request from a Qwen profile preserves the intended candidate while selecting a suitable fictional demonstration model. Review both labels before choosing Run sample. The candidates’ 8,192-token context and 1,024-token output values are proposed pilot limits, not measured capabilities. Any timing, usage or Work Credit amount attached to the sample remains illustrative.
Use the receipt to decide what to evaluate next
Inspect the project, starter, intended candidate, actual fixture and completion state in the local record. A connected release must make the corresponding execution evidence available without substituting a familiar model name for verified identity. Exported samples should retain their simulation labels. Use this preparation to define a bounded test; no live model request, code execution, GPU allocation or payment occurs.
Service maturity
Explore the workflow today
Three guided starters prepare browser-local samples with distinct intended-candidate and fictional-fixture labels.
Before connected service access
Connected inference awaits qualified SAVRN-operated deployments, approved access, tested request contracts and reconciled metering.
What this makes possible
The intended outcome is a clear request trail that helps developers reproduce an experiment and helps project owners understand its resource use.
Build toward the work that matters.
Tell SAVRN what your institution needs to run, who reviews the results and where its data must stay. That workload defines the next service to qualify.
Discuss an AI project