Release notesGelim Account release notes
On this page
Release notes

Gelim Account

6 November 2026

  • Models keep working when a price page changes: on 6 November no model could run from midnight (UTC) until about 10:00, because Google had moved its price page and Gelim Account couldn't renew prices. Now a model keeps running at its last known price while a page can't be read, each provider's prices are renewed on their own, and Google's new page is read.
  • Search and deep research keep working the same way: they run at their last known price while the price page can't be read, so a change on that page no longer stops them.
  • Newer transcription: recordings are transcribed with the latest transcription model, with better speaker separation and language switching. Prices and codes you say come out as words, such as "15 dollars and 20 cents". What you pay doesn't change.
  • Clearer when models can't run: if Gelim Account can't run models for a while, Gelim tells you that, instead of saying the model you chose isn't available.

27 October 2026

  • Every model works again: from midnight (UTC) until 09:30 on 27 October, no model could run, so live translation, chats, actions and the model pickers stopped working. A change on OpenAI's price page had stopped Gelim Account from renewing prices for a week. Prices are read correctly again, and kept up to date.

20 October 2026

  • Gemini 4 Argon is in your models, at Google's published price.

19 October 2026

  • Files from your chats: ask for an Excel sheet, a Word report, slides, a PDF or a web page, and it arrives as a card in the chat, on the Mac (2.1.8 and later), the web and the iPhone. Files from your chats
  • Your files kept safer: files and images from your chats move to Gelim Account's own encrypted storage on our server, backed up every night. Nothing changes in how you use them.
  • Bigger files: files your chats make can be up to 50 MB (was 20 MB), and a large download picks up where it stopped.
  • Deleted means gone: when you delete a chat for good, its files leave our storage and backups within about 9 days.

18 October 2026

  • Models, at a glance: Settings → Models is one table you can sort and filter, with each model's speed (first token and tokens per second, for text and for pictures), its price per million tokens and its context size.
  • New models within the hour: a new model from OpenAI, Anthropic or Google appears as soon as its price is published, marked New for 14 days, and its speed is measured soon after.
  • Your picker: choose which models your apps' pickers show with Show in the picker. New accounts start with the newest Claude Opus, Sonnet, Haiku and Fable, each GPT tier, and Gemini Pro, Flash, Flash-Lite and Argon.
  • Claude Haiku 5.5 is available.
  • Long answers from OpenAI models no longer stop partway.

Next steps

Updated 6 November 2026