Tuning Omni AI for cost and quality

Two sets of controls decide how much Omni’s AI costs and how well it answers:

LLM settings in the AI Hub, which set the LLM and thinking level for your whole organization
The ai_settings parameter in a model file, which controls validation, context pruning, and query scope for the Omni Agent on that model

Together they form a spectrum between “cheap and fast” and “accurate but expensive.” You can tune Omni for the workloads it actually serves instead of accepting defaults everywhere. This guide maps those knobs to three starting profiles — cost-optimized, balanced, and max-quality — that you can adopt directly or adapt to your organization.

Requirements

To follow this guide, you’ll need:

Organization Admin permissions to change LLM settings
Connection Admin or Modeler permissions on the model whose ai_settings you want to edit
An understanding of the parameters documented on the ai_settings reference

How settings affect cost

Every AI turn in Omni costs LLM tokens. The table ranks how much each setting can change that cost. It doesn’t mean that a setting is expensive on its own.

Setting	Cost impact
Default LLM	High. Organization-wide. The most capable LLMs are multiple times more expensive per LLM token than mid-range LLMs.
Thinking level	High. Organization-wide. Each level adds reasoning LLM tokens to every turn; High adds the most. The organization thinking level applies to an LLM a user picks, too.
Approved LLMs	Varies. Organization-wide. Users can pick any approved LLM for a chat, so approving a more expensive LLM lets users choose to spend more.
`validate_analysis`	High. Adds LLM requests per analytical turn, each on the same LLM as the chat.
`conversation_prune_length`	Medium. Claude LLMs only. Higher thresholds mean more LLM tokens billed per turn on long conversations.
`query_all_views_and_fields`	Low–medium. A larger search surface means more views considered per query, which costs LLM tokens.

Light tasks, such as summaries and chat names, always run on a lower-cost LLM with no thinking. You don’t need to tune them.

Tuning profiles

Each profile combines organization-wide LLM settings with an ai_settings block for your model files. You can mix settings across profiles and use a different ai_settings block on specific models when individual use cases warrant it. See Mix and override profiles below.

Profile	Best for	Tradeoffs
Cost-optimized	High-volume AI usage, recurring questions against a well-curated topic set	Smaller query surface, no query validation, earlier context pruning, no extended reasoning
Balanced	Most organizations, most of the time. Mirrors Omni’s default behavior.
Max quality	Highly complex questions and datasets, optimizing for depth over speed, debugging accuracy issues with default settings	Potential for higher costs due to query validation and usage of the most capable LLM with a high thinking level

Cost-optimized

This profile is best for high-volume AI usage where per-query cost matters more than marginal accuracy, or models where end users are doing bounded, familiar analysis. For example, recurring questions over a well-curated topic set. Click the tabs to see how to configure the LLM settings and Model file.

LLM settings
Model file

Define the following in the LLM settings page:

Setting	Value
Default LLM	An LLM whose description in the LLM list says it balances speed and capability, not the most capable one
Thinking	Off
Approved LLMs	Remove approval from the most capable LLMs, so users can’t pick them

Add the following block to your model file:

Cost-optimized ai_settings profile

ai_settings:
  query_all_views_and_fields: disabled
  validate_analysis: disabled
  conversation_prune_length: short

The tradeoffs of this approach are:

A smaller query surface (topic-scoped only), so ad-hoc views aren’t available to the AI (query_all_views_and_fields: disabled)
No self-correction before users see results (validate_analysis: disabled)
Earlier context pruning on long chats (75,000 LLM tokens vs. 175,000 at the default) (conversation_prune_length: short)
No extended reasoning before each answer (Thinking set to Off)
Users can’t move a hard question to a more capable LLM

Balanced (default)

This profile is best for most organizations, most of the time. It is the closest to Omni’s out-of-the-box behavior. It differs from the defaults in one place: conversation_prune_length: max instead of long, which opts you into a larger context window on an Omni-managed 1M-context Claude or GPT model. You can copy the ai_settings block into your model to make the intent explicit. Click the tabs to see how to configure the LLM settings and Model file.

LLM settings
Model file

Define the following in the LLM settings page:

Setting	Value
Default LLM	The Deployment default option
Thinking	Default
Approved LLMs	All of the LLMs that Omni hosts

Add the following block to your model file:

Balanced (default) ai_settings profile

ai_settings:
  query_all_views_and_fields: enabled
  validate_analysis: disabled
  conversation_prune_length: max

This approach is the default because:

Omni’s default LLM is strong enough for most analytical tasks without the pricing of the most capable LLMs
Users can still pick a more capable LLM for a hard question, one chat at a time
conversation_prune_length: max preserves context while opting you into a larger context window on an Omni-managed 1M-context Claude or GPT model

Max quality

This profile is best for highly complex questions and datasets, and when you want to optimize for exploration depth and answer quality over speed. It can also be useful when you’ve seen accuracy issues with the defaults and want to raise the floor before debugging further. Click the tabs to see how to configure the LLM settings and Model file.

LLM settings
Model file

Define the following in the LLM settings page:

Setting	Value
Default LLM	The most capable approved LLM
Thinking	High
Approved LLMs	The most capable LLM, with Test only cleared

Add the following block to your model file:

Max quality ai_settings profile

ai_settings:
  query_all_views_and_fields: enabled
  validate_analysis: enabled
  conversation_prune_length: max

While this approach can yield higher quality results, there are a few tradeoffs:

validate_analysis: enabled adds LLM requests per analytical turn.
The most capable LLM with Thinking set to High is the single biggest cost multiplier. Use it only if the accuracy lift is worth it for your workload.
The default LLM and thinking level apply to your whole organization, including the features that have no LLM picker. You can’t limit them to one semantic model. See How Omni selects an LLM.

Mix and override profiles

The profiles above are starting points — you don’t have to adopt one in its entirety. A few common compositions:

Mostly default with quality insurance. Start from Balanced and flip validate_analysis to enabled. This adds self-correction without upgrading the LLM or thinking level. It adds LLM requests to each analytical turn, but costs less than moving the whole organization to a more capable LLM with a higher thinking level.
ai_settings: validate_analysis: enabled
Balanced default, most capable LLM on request. Keep the Balanced LLM settings and make sure the most capable LLM is approved. Add a description that tells users when to pick it, such as “Use for complex, multi-step analysis.” This concentrates spend on the questions that need it, without paying for it on every turn.
Cost-optimized with fuller context. Start from Cost-optimized but use conversation_prune_length: long instead of short if your users tend to have long chat sessions where losing earlier context would hurt more than the LLM token savings.
ai_settings: query_all_views_and_fields: disabled validate_analysis: disabled conversation_prune_length: long

Apply a profile

Navigate to AI Hub > General and click the LLM settings tab.

Set the Default LLM and Thinking values first, then adjust approvals. You can’t remove approval from the LLM that is the current default. See LLM settings for more information.If you aren’t sure which profile to use, start with Balanced and adjust only when you see a specific cost or accuracy issue.

To try a new LLM first, approve it as test only and use an eval run to compare its accuracy and credit usage with your current default.

Navigate to the model IDE and open the model settings (model) file.

Add the ai_settings block from the profile you want to use to the model file.

Save your changes and promote them to the shared model.

Open a new chat and confirm that the LLM picker lists the approved LLMs with the default marked Default.

Monitor and iterate

After applying a profile, check the Analytics > Credit tracking dashboard periodically to see how your users are interacting with Omni AI. Look for:

Question volume and types — Helps you judge whether the profile still fits the actual workload
Turn counts per conversation — Many long conversations mean pruning is more likely to drop earlier context. Consider raising conversation_prune_length, or investigate why conversations run long

Start balanced, upgrade a single dimension (such as validate_analysis), observe the change, and continue from there rather than jumping straight to max quality.

Next steps

AI credit tracking — Track your organization’s AI usage over time
LLM settings — Approve LLMs and set the organization default
ai_settings reference — Full parameter documentation
Optimize models for Omni AI — Curate the context Omni AI sees, which can help improve quality instead of moving to a more expensive LLM
Learn from conversation — Use real user questions to improve the model over time