Get started
First inference walkthrough
Choose a prepared Qwen candidate and one of three guided starters, then review a small public or invented input before running a browser sample. The walkthrough explains supplied-evidence boundaries, null values for missing extraction fields and code suggestions that remain unexecuted, followed by a receipt identifying the actual fixture.
SAVRN Cloud · Coming soon · Public preview · Browser-local simulation · No live compute or payments
Choose a candidate and a task
Start in Compare models, inspect a candidate profile and choose Prepare request. The Playground keeps the intended candidate visible while selecting a fictional fixture for the demonstration. You can also begin directly with one of the three guided starters. Connected Qwen execution is Coming soon; this walkthrough sends no model request.
Confirm the active project before entering public or invented material. Candidate profiles propose 8,192 context tokens and 1,024 output tokens for a future pilot. These are not measured service limits, and separate demo controls describe only the local simulation.
Evidence review
Supply short excerpts with labels such as Source A and Source B, then ask a bounded question. Request supported claims, references to the supplied passages and a list of unresolved points. Treat absent evidence as absent. A URL is not fetched automatically, and the deterministic sample is not a factual verification of the supplied source.
Structured extraction
Name the fields you want, then provide a small text. Ask for null when a value is missing and for uncertainty to remain explicit. For example, extracting an organization, date and project status should not invent a date merely because the output shape includes one. This is a request-writing exercise; no schema-enforcing Qwen capability has been qualified.
Research code
Describe a narrow analysis task or supply a code fragment. Ask for assumptions, an explanation of the proposed change and checks a researcher could perform. Review the response as a sample artifact. The preview does not execute supplied code, generated code, shell commands or tests.
Run and inspect the labeled sample
Choose Run sample, then inspect its completion state and receipt. Identify the intended candidate, actual fictional fixture, starter, project and sample usage. Open Usage to follow the record. A sample Work Credit amount is illustrative, not a bill or a current quote.
View API shape provides explanatory JSON. Preview Keys do not grant real service access. See API publication requirements for the endpoint, compatibility, authorization and accounting evidence a connected quickstart must eventually provide.
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