Experimental Features on Skytells APIs
Limited-access experimental features and model rollout policy.
Experimental Access
At Skytells, research is a core strategic priority. We invest 30% of our resources in research and development, and as part of this commitment AI Research, we may release experimental capabilities to approved accounts in limited phases.
Experimental features help us validate new capabilities with real-world feedback before broad release. This page defines what experimental access means, who can request it, and how these features are governed.
What Experimental Means
Experimental features are pre-general-availability capabilities that may include:
- Beta models
- Early-access APIs
- Preview parameters or response fields
- New performance modes or infrastructure paths
These features are intended for evaluation, prototyping, and controlled production pilots. They are not guaranteed to be stable or permanently available.
Access Eligibility
Access is limited to selected accounts with prior approval.
Eligibility is typically based on:
- Use-case alignment with the experimental scope
- Safety and policy readiness
- Expected request volume and infrastructure impact
- Region and compliance constraints
- Feedback quality and collaboration readiness
Approval is feature-specific. Access to one experimental model or endpoint does not automatically grant access to others.
Lifecycle and Stability
Experimental features move through iterative phases:
- Internal validation
- Limited external rollout
- Expanded preview (if successful)
- General availability, redesign, or retirement
During this lifecycle, we may change or remove functionality, including:
- Model versions and behavior
- API request and response schemas
- Rate limits and quota policies
- Availability windows and regions
For production systems, always implement graceful fallback logic and version-aware integrations when using experimental features.
Usage Expectations
If you are approved for experimental access, you are expected to:
- Use features only within approved use cases
- Respect platform policies and applicable laws
- Avoid relying on undocumented behavior
- Monitor error rates and quality outcomes
- Share actionable feedback with reproducible examples
Known Limitations
Experimental capabilities may have one or more of the following limitations:
- Lower uptime targets than generally available services
- Limited observability or incomplete metrics
- Stricter rate limits
- Temporary model IDs or naming changes
- Breaking changes with short notice
How to Request Access
To request access, contact support with your account details, intended use case, and expected volume. Our team will review eligibility and share the next steps.
Include the following in your request to accelerate review:
- Account or workspace identifier
- Target model or feature name
- Use-case summary and expected users
- Estimated requests per minute/day
- Region and compliance requirements
- Desired start timeline
Support and Change Communication
For approved accounts, updates may be communicated through documentation updates, support channels, and direct rollout notices when applicable.
If a feature is changed, paused, or retired, we recommend switching to a stable generally available alternative as soon as possible.
How is this guide?