Model SelectionCopy Anchor Link

Claira routes each scan through an AI model you select for the case. The model choice affects latency and how the model handles difficult documents. It does not change what a scan costs: your plan is charged the same tokens per document whichever model you choose. This page explains where to change the model and how to validate the choice before scaling up.

Where to choose the modelCopy Anchor Link

Model selection is configured in the Claira Settings area for the active case.

1. Open your case workspace and select the Claira pane.
2. Open the Modes menu and choose Settings.
3. Under Your settings, click Model settings to open the section. While it is closed, its heading already shows the model you are using.
4. Select a model in the Model dropdown. The choice is saved straight away and applies to your own scans in this case, so if your team should all use the same model, agree on it together.

Available modelsCopy Anchor Link

Claira offers region-resident model options. On a Claira Private case, the case runs on your own model instead, and this picker is not shown.

• Gemini Flash 3.5 (the default everywhere). The newest Gemini Flash generation, with a large context window that suits long documents. Runs inside your tenant's home region. For complex, ambiguous, or borderline documents, raise its Reasoning level rather than switching model (see Model Settings).
• Claude Sonnet 4.6 (Australia tenants only). A Claude model, hosted inside the Australian data residency boundary. A strong alternative when the Gemini models are busy, and a second opinion from a different model family on borderline documents. Text scans only (see Which models can run a media scan). Does not appear in the picker for Canada tenants.
• Amazon Nova Lite (Canada tenants only). A fast, lightweight model hosted inside the Canadian data residency boundary. Useful for quick first-pass triage or as a second opinion from a different model family. Text scans only. Does not appear in the picker for Australia tenants.
• GPT-OSS 120B (Canada and Australia tenants). An open-source model hosted inside your own data residency boundary. Useful when your review protocol calls for an open-weights model alongside the Gemini options, or when you want a second opinion from a different model family on a benchmark set. Text scans only. Does not appear in the picker for Germany tenants.
• MAGPiE (Australian sovereign) (Australia tenants only). A model built and hosted by an Australian-owned company on its own hardware in Australian data centres, so both the data and the operator stay onshore. Useful where a review protocol or procurement requirement calls for an Australian-owned provider rather than an Australian region of a global one. Text scans only.
• Llama 4 Maverick (Australia tenants only). Hosted in Australia. The fastest option in the Australian picker and the one to reach for on high-volume first-pass coding. Runs Image scans of PNG, JPEG and WebP images and single-page TIFF or JPEG 2000 scans, but not PDFs, audio or video (details). Produces shorter answers than the other models, good for coding and classification, less suited to long summaries.
• Gemma 4 31B (Australia tenants only). A lightweight open-weights model hosted in Australia. Runs Image scans of PNG, JPEG and WebP images and single-page TIFF or JPEG 2000 scans, but not PDFs, audio or video (details). Fast once warm, but the first scan after a quiet period can take noticeably longer.
• Qwen 3.8 Max (Germany tenants only). Hosted inside the German data residency boundary. The largest context window Claira offers in Germany, useful for long documents that the Gemini Flash models truncate. Runs Image scans of PNG, JPEG and WebP images and single-page TIFF or JPEG 2000 scans, but not PDFs, audio or video (details). Slower per document than the Gemini options.
• GLM 5.2 (Germany tenants only). Hosted in Germany. A very large context window at lower latency than Qwen 3.8 Max. Text scans only, and it produces shorter answers.

Retired modelsCopy Anchor Link

Gemini Flash 2.5 and Gemini Pro 2.5 were retired in October 2026, when they reached the end of their life. If you or a teammate had chosen one, Claira switched that choice to Gemini Flash 3.5 for you. A bulk task that was already running when the change arrived finished on the model it started on; any task you start or resume now runs on Gemini Flash 3.5. Your prompt, fields, temperature, Top P and reasoning level were kept. Scans and tasks that already finished still show the model that answered them.

Switching between the models above (where available) is a model-name change only: your prompt, fields, and case settings carry over unchanged. Your plan is charged the same per document whichever model you choose.

To run a case on your organization's own model, activate it on Claira Private instead of a standard plan. A case on a standard plan always uses the models in this picker.

Where each model runsCopy Anchor Link

Every model Claira serves carries a small flag in the picker showing the country it runs in: the Canadian, German, or Australian flag depending on your tenant. Whichever of those you choose, your documents are processed inside your own data residency boundary; the flag is there so you can confirm that at a glance without leaving the settings screen.

Every model Claira serves is held to the same model provider policies: no training on your content, no retention, and no sharing. For the providers that host the models in your region, contact support@claira.to.

Automatic model routingCopy Anchor Link

An AI model is sometimes busy: it is over capacity, or it stops responding for a while. Claira always retries a busy model a few times first. If the model still does not answer and Automatic model routing is on, Claira answers with another approved model in your country instead of telling you that the AI model is busy.

Turning it on or offCopy Anchor Link

1. Open the Modes menu and choose Settings.
2. Under Your settings, click Model settings.
3. Turn Automatic model routing on or off. It sits beside the Model picker, and your choice is saved straight away.

Like your model, the setting belongs to you and to this case: turning it on does not change it for your colleagues, or for your other cases. It starts on in cases created after this feature was released, and off in cases that already existed, so a review already under way only changes model if you decide it should. You can change it at any time. If you do not see the switch, automatic model routing is not available on your deployment yet.

What happens when a model is busyCopy Anchor Link

• Claira tries the other models available to you, in a fixed order, until one answers. It keeps trying for up to about a minute. If every model is busy, you see the same error as before.
• Nothing appears in the panel when Claira switches: your answer arrives as usual.
• The record of each scan and each chat reply keeps the model that actually answered. Wherever a record shows a model, for example an exported history, it shows that model.
• Your organization's admins can see each switch in the Admin Portal: which feature it was, which model was busy, which model answered and, for a bulk task, how many documents. See Activity.
• A switch does not change what you pay: your plan is charged the same per document whichever model answers.

Your documents stay in your countryCopy Anchor Link

Automatic model routing only uses models already offered in your Model picker, all held to the same model provider policies. It never sends your documents outside your deployment's country.

Where it appliesCopy Anchor Link

It covers every AI feature: single scans, bulk tasks run in your browser, Multi-Code, Agent Mode, Case Context, Insights, Early case assessment, the Prompt Generator and Quickstart, Revise with Agent, and the names Claira gives your chats and bulk tasks.

It does not apply to:

• Claira Private cases. A model you connect is never swapped for another one, and Claira never moves work onto it.
• Bulk tasks run with Background Processing. They keep waiting and retrying until the model answers, as before.

Which models Claira moves toCopy Anchor Link

Claira only moves to a model that can do the job:

• A bulk task only moves to models suited to large runs.
• An Image, Audio or Video scan only moves to models that can read that file.
• If you picked a model from a sovereign provider, Claira only moves to other models from the same provider.
• Agent Mode and Case Context only move to models that support what they need.

If none of those models is free, you see the busy error as before.

Claira Private casesCopy Anchor Link

On a Claira Private case, Model settings has no Model picker and no Automatic model routing switch. The case runs on your own model: the shared connection for everyone on the case, set up under Team settings, or your own local model while you have one turned on, chosen in the Your model section. Nothing on the case is ever sent to one of the models listed on this page.

To move a case between a standard plan and Claira Private, deactivate it and activate it again. See Choosing Claira Private.

Saving and resettingCopy Anchor Link

Model settings save as you change them. There is no Save button: the bar at the bottom of Settings shows Saving your changes while a change is on its way and All changes saved once it is stored. If a change cannot be saved, the bar says Your changes were not saved, gives the reason, and offers Try again. Done takes you back to Prompt Lab.

Only the model, temperature, Top P, reasoning level and automatic model routing are saved from this section. Switching between modes saves the mode you are in, and sends any change that is still on its way, but nothing else. The prompt you write in Prompt Lab and Bulk Scanning, and your Multi-Code setup, save on their own as you type: the row with Save Prompt shows Unsaved changes, Saving your changes and All changes saved in the same way, and after a reload you find them as you left them. Save Prompt still adds the prompt to your prompt history. If the prompt cannot be saved on its own, the row says Your changes were not saved with the reason and asks you to click Save Prompt, which saves it a different way; your text stays on screen either way. Your Scan as choice, the way you last ran a bulk scan and your default chat visibility are remembered too.

Reset model settings puts the model, temperature, Top P and reasoning level back to their defaults, and saves them. Your prompt, your connected fields and your Multi-Code setup stay exactly as they are. The button is unavailable while everything is already at its default.

Models and Background ProcessingCopy Anchor Link

Background Processing always runs on a Gemini model, regardless of the model selected in Settings. If a non-Gemini model (such as Claude Sonnet 4.6) is selected when you start a background task, the confirmation dialog shows a note that the background run may use a different model, and the task's history records the model that actually ran. Foreground bulk scans are unaffected: they use exactly the model you selected.

Before you change models mid-reviewCopy Anchor Link

• Confirm field mappings still match your prompt output. The model does not change your fields, but different models may phrase answers slightly differently. Re-run a small benchmark set after switching.
• Re-test 10 to 25 documents with Single Review before running a new Bulk Scan.
• Token cost is not a factor. A scan costs the same tokens on every model, so compare models on quality and speed.

Model settings (temperature, Top P, reasoning)Copy Anchor Link

After you pick a model, you can tune behavior with Temperature, Top P, and Reasoning level. See Model Settings for defaults and task-based recommendations.

If results change after switching modelsCopy Anchor Link

1. Run the same prompt on the same benchmark documents you used before the switch.
2. If outputs drift, tighten the prompt (definitions, allowed values, output format) before blaming the model.
3. If only one model fails with Error Codes such as E-AI, try the other available model and contact support if both fail.

RelatedCopy Anchor Link

• Understanding Results
• Troubleshooting

Need help? Contact us at support@claira.to.