Changelog
Every release, in order. Model additions land continuously and are noted here in the fortnight they ship.
2.4
Auto-Route v3 and per-seat caps
- NewAuto-Route v3 uses your workspace's own feedback history to score models per task class, rather than a global average. Early workspaces are seeing another 8–12% off routed spend.
- NewPer-seat spend caps on Team, with a choice of soft (warn) or hard (refuse) enforcement.
- BetterSide-by-side now streams all four columns in parallel rather than in pairs. A four-model comparison finishes in the time the slowest model takes.
- FixThreads pinned to a specific model no longer lose the pin when a comparison is run inside them.
2.3
Prompt library versioning
- NewLibrary prompts are versioned. Every edit creates a revision with a diff, and any revision can be restored.
- NewPrompts can be pinned to the model they were tuned against, so Auto-Route leaves them alone.
- BetterUsage analytics gained a per-member breakdown with cap status.
- FixCSV export no longer truncates workspaces with more than 10,000 rows in a period.
2.2
Region pinning and SCIM
- NewEnterprise workspaces can pin processing to the United States, European Union or Australia.
- NewSCIM 2.0 provisioning for Okta and Entra ID.
- BetterModels with no endpoint in a pinned region are now hidden from the catalogue instead of failing at request time.
- FixSSO sessions no longer expire mid-stream on long reasoning responses.
2.1
Credits that roll over
- NewUnused credits roll over for 60 days and are drawn oldest-first.
- NewAuto top-up, off by default, with a configurable threshold.
- BetterInvoices now itemise spend by provider.