> ## Documentation Index
> Fetch the complete documentation index at: https://help.nops.io/llms.txt
> Use this file to discover all available pages before exploring further.

# AI Model Version Upgrade

> Switch to a newer model version in the same family and tier that costs less per token.

AI providers regularly release a newer version of a model at the same capability tier, often at a lower price per token. nOps finds people, workloads, API keys and projects still running an older version when a newer one in the same family and tier is cheaper and already in use in your organization, and estimates the saving from moving the same work to it.

| | |
| - | - |
| **Cloud** | AI |
| **Category** | AI |
| **Applies to** | Claude Enterprise, Anthropic API, OpenAI API, Amazon Bedrock, Gemini API |
| **Action** | Modify |
| **Effort** | Low |
| **Restart needed** | No |
| **Can be rolled back** | Yes |
| **Analysis window** | Last 30 days |
| **Source** | nOps |

## Why it matters

Older versions stay in configuration files, gateways and user defaults long after a newer, cheaper one is released. Because the tier is the same, the move costs you no capability.

## How nOps finds it

nOps reads the last 30 days of AI billing per owner and model: people and workloads on AWS Bedrock (Claude, and OpenAI models sold through AWS Marketplace), Claude Enterprise users, Anthropic API keys and OpenAI API projects. A target must be the same family and tier with the same or a newer version, and must be a model your organization spent at least $100 on in the window. The owner must have spent at least $25 on the current model, and the target must have a price for every token type the owner used. The saving is each token type's cost times the difference between the two models' rates on your own bills, and must be at least \$25 a month. When both a version upgrade and a cheaper tier are possible, nOps recommends the version upgrade.

## Before you start

The ability to change the model ID the workload, API key or user calls.

## How to implement

1. Find the owner, current model and suggested model in the recommendation.
2. Check the suggested version on a sample of real prompts. Newer versions can format answers or call tools differently.
3. Change the model ID in the application configuration, the gateway (for example a LiteLLM model alias), or the user's default model.
4. For Bedrock, update the model or inference profile ID in your code and confirm the IAM role is allowed to invoke the new model.
5. Update any Terraform, CloudFormation or deployment configuration that pins the old model ID.

## Risks and considerations

* A newer version can behave differently even at the same tier. Run your evaluations or a short side-by-side test first.
* Older versions are eventually retired by the provider, so moving early also avoids a forced migration later.
* To roll back, switch the model ID back.

## FAQs

> [accordion] How is this different from Model Tier Right-Sizing?
>
> A version upgrade stays at the same tier, so you keep the same capability. Tier right-sizing moves to a less capable tier.

## Related

* [AI Model Tier Right-Sizing](https://help.nops.io/inform/supported-recommendations/ai-model-tier-right-sizing)
* [AI Model Cheapest Tier](https://help.nops.io/inform/supported-recommendations/ai-model-cheapest-tier)
* [Supported recommendations](https://help.nops.io/inform/supported-recommendations.md)


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.