What has changed
Mistral released Large 4 in public preview on 6 October 2026. Its official changelog describes a general-purpose multimodal model with a mixture-of-experts architecture and a one-million-token context window. The API preview is available now; the model weights are still forthcoming. “Open-weight” therefore describes the intended release, not an immediately available self-hosted download.
The preview is a useful testing opportunity, but claims about leading performance should be treated as Mistral’s claims until they are reproduced on the work a business actually needs.
What Australian businesses should test
A larger context window can help a team work with long documents, but capacity alone does not establish accuracy. Build a small test set with known answers, including a difficult case where the correct response is to admit uncertainty.
- Compare accepted results, response time and the cost of the complete task.
- Test document citations and whether answers remain accurate across long inputs.
- Check processing location, retention, access controls and contractual terms.
- Keep human approval for actions that affect customers or external systems.
The practical takeaway
Add Large 4 to an existing comparison rather than switching a production workflow on launch day. Technical teams interested in self-hosting should wait for the weights and inspect the released licence and deployment requirements. For most small businesses, the useful question is whether it completes a defined task more reliably or economically than the tool already in use.
Source checked: Mistral’s official changelog, 7 October 2026.