Every chat remembers its model
The model picker in the top-left used to be a single global setting — change the model in one chat and every other chat changed too, so returning to an older conversation showed whatever model you last picked elsewhere, not the one that conversation was actually using. Now every session remembers its own model: open a chat and the selector restores the model and provider it was last using, and switching between chats switches the model along with them. New chats still start from your current default, and if a session's remembered model is no longer available — provider disconnected, model retired — the selector quietly falls back instead of stranding you on a dead model. The choice is persisted on every turn, including edit-and-resend, so it survives restarts.
No more false "can't read images" warnings
The up-front vision check added in 1.2.1 was over-eager and flagged some models that can read images. Attachment support is now gated by a curated allowlist of vision-capable model families, validated against the live models.dev catalog, so capable models are no longer wrongly warned at attach time.