Release Notes
A full rundown of TensorX releases, improvements and fixes.
Model changes
We are retiring Qwen3.8-Flash-Next because of Qwen licensing changes and lower uptake, and reallocating GPU capacity to more widely used models. This round has a one-off shorter notice period of one to five days; our standard remains two weeks. Old model IDs keep working and are routed to their replacement automatically. Rerouted requests are billed at the replacement model's rate.
qwen/qwen3.8-flash-nextis retired from 31 October 2026 and routes toqwen/qwen3.8-27b.minimax/minimax-m3is retired from 31 October 2026 and routes tominimax/minimax-m3.1-flash.deepseek/deepseek-v4.1-flashis now served bydeepseek/deepseek-v4.1-flash-1017(from 29 October 2026). Same price.moonshotai/kimi-k2.7-codeis retired from 4 November 2026 and routes tomoonshotai/kimi-k3.
See Model Deprecation for the full list.
Billing & Invoices
Users can now add specific email addresses to receive communications relating to billing - Invoices initially. These emails will be sent when an account top-up (transaction) is made. Invoices will still remain accessible via the dashboard. To enable this email feature: Billing > Manage Billing > Billing Emails (base of details form) > Update
Archiving of keys when joining a team
When a user joins a team, their existing keys will be archived. These keys will reactivate upon leaving a team. We have added clearer alerts and messaging on the platform to convey this.
New modal details on this flow
Appended elements on the usage table on the dashboard
Last updated
Was this helpful?

