Skip to main content

AI model

Who can do this

Administrators with the AI permission.

Axiome runs its AI features — Axi conversations, dashboard building, analysis — on the model of your choice, picked from the current lineups of three providers. The choice applies to the whole workspace.

Settings → Admin → AI

Choose the model that powers AI features across your workspace.

The lineup

Models are grouped by provider, from each provider's fastest tier to its flagship. The $ marks show relative cost per conversation, from $ (cheapest) to $$$$ (premium).

Anthropic

ModelCost
Claude Haiku 4.5$Fast and inexpensive — good for quick questions, less thorough on complex dashboard building
Claude Sonnet 5$$Balanced, tuned for reliable analysis and dashboard building
Claude Opus 5$$$Stronger reasoning for demanding analyses
Claude Fable 5$$$$Anthropic's most capable model — the deepest analyses, at a premium price

OpenAI

ModelCost
GPT-5.6 Luna$The cost-efficient tier — quick tasks at minimal cost
GPT-5.6 Terra$$The default — the balanced model for everyday analysis and building
GPT-5.6 Sol$$$OpenAI's flagship — the strongest reasoning in the GPT family

Mistral AI

ModelCost
Mistral Large 3$The European provider's value workhorse — capable and inexpensive
Mistral Medium 3.5$$Mistral's open-weight flagship, strong on agentic work

Click a model to select it. The change applies immediately to new messages, for every member and every AI surface — there is nothing for your team to do, and conversations in progress simply continue on the new model from their next message.

Picking

All models drive the same tools; they differ in depth, speed and price. If you have no preference, keep the default — it is the configuration Axiome tunes against. Premium tiers (Opus, Fable, Sol) buy deeper analysis on demanding questions; cheap tiers (Haiku, Luna, Large) trade some rigor for speed and cost.

Bring your own API key

By default, AI runs are billed to Axiome. If you prefer them billed to your own account with a provider, add your own API key under Settings → Admin → AI → API keys — one key per provider (Anthropic, OpenAI, Mistral AI).

  • The key that matters is the one for the selected model's provider: pick an OpenAI model and your OpenAI key is used; the others sit idle until you switch.
  • When you save a key, Axiome verifies it with the provider first — a mistyped or revoked key is rejected on the spot.
  • Keys are write-only: they are stored encrypted and never displayed again. The page shows only the last 4 characters so you can tell keys apart. Replace or remove a key at any time.
  • Remove a key (or leave a provider unconfigured) and runs simply bill to Axiome's key again — nothing breaks.

If a stored key later stops working — revoked, or out of funds — AI messages fail with a clear notice naming the workspace key, and any admin can update it from this page.

Runs billed to your own key don't consume members' AI credits.

What runs on your key

Your key covers the AI conversations and dashboard building powered by the selected model. Axiome's built-in background features (AI summaries, alert analysis) stay on Axiome's own keys.

What the choice does not affect

  • AI summaries and alert analysis stay on the model Axiome pins for them, so shared summaries remain consistent for everyone and cached results stay valid.
  • Past conversations keep the record of the model they ran on; nothing is rewritten.

Usage and credits

AI runs are metered per model — the Usage page breaks credits and tokens down by model and provider, so you can compare models after a switch. Pricier models consume a member's monthly credit allowance faster.