Skip to content
AskVault Docs
Try Free →

Model selector

Written by , Founder, AskVault · Reviewed by Aashiq
Last updated: · 3 min read

Default models per plan

Auto-assigned:

  • Free. Lightweight model. Fast and capable.
  • Starter, Growth. Mid-tier. Better reasoning.
  • Business. Higher-capability. Best for complex queries.
  • Enterprise. Custom or bring-your-own.

Override on Business

For Business workspaces:

  1. Settings > Model Selector.
  2. Pick from available list.
  3. Save.

Override applies within 30 seconds to new conversations.

Bring your own model (Enterprise)

For Enterprise customers:

  • Self-hosted open-source models. Llama, Mistral, etc.
  • Private cloud deployments (your AWS/GCP).
  • Custom fine-tuned models.

Configure via Enterprise onboarding.

Comparing models

Use Compare Models in Chat Playground:

  • Run same query against 2 to 3 models.
  • See quality, latency, cost differences.
  • Pick based on your priority.

Cost vs quality trade-offs

Roughly:

  • Lightweight (Free tier). Fastest, lowest cost per token. Good for simple FAQs.
  • Mid-tier (Growth). Better at multi-step reasoning.
  • High-capability (Business). Best for complex queries, long context.

For high-volume bots: prefer mid-tier unless quality requires higher.

Per-skill model override

Some skills can use different models:

  • subscription_manager can pin a deterministic model for billing accuracy.
  • content_recommender can use a faster model for low-stakes suggestions.

Configure under each skill's Advanced > Model.

Model versioning

When AskVault rolls out a new version:

  • Default model auto-updates to latest unless pinned.
  • Pin specific version on Business and above for stability.
  • Notification 30 days before retiring old versions.

Limits

  • Available models. Plan-dependent.
  • Override granularity. Workspace or per-skill.
  • Model change rate. Up to 10 per day (anti-thrashing).

Common pitfalls

Switched to a lower model, quality dropped. Revert in 30 seconds; or fine-tune the system prompt for the lower model.

Latency jumped after model switch. Higher-capability models often slower. Trade-off.

Cost spike. Higher-tier model means more tokens per query. Audit usage post-switch.

FAQ

Can I see which model is currently active?

Yes under Settings > Model Selector > Status.

Will switching models lose conversation context?

No. Past messages preserve; the next turn uses the new model.

Does switching affect cost immediately?

Yes. Next query bills at new model rate.

Was this page helpful?