Skip to main content
​

Product

  • Email sign-in shows why an attempt failed, such as a wrong password or an unverified email, on the sign-in page instead of a generic error page (#116).
  • Email sign-up lists the password rules (lowercase, uppercase, number and symbol) and checks them before the form is sent (#113).
  • The models table shows prices per million units correctly and uses the full page width; before, some rows showed extremely high prices (#113, #116).

API & SDKs

  • Gemini 3 requests map reasoning.effort to the thinking levels each model supports, from minimal to high (#113).
​

API & SDKs

  • Breaking: /v1/responses/ws is now marked experimental and rejects plain model slugs; send openai/<model> IDs, and use POST /v1/responses for production (#112).
​

Product

  • Model pages show one provider logo in light mode; before, the light and dark versions could both appear (#81).

API & SDKs

  • Requests to Baseten, Chutes, Cohere, Together and Cerebras drop parameters the provider doesn’t support instead of failing with an upstream 400 (#82).
​

Product

  • Settings has a sidebar with Account, Billing, Usage and API sections split into tabs, and Account has separate details, MFA and danger zone pages (#54, #55).
  • Settings -> Credits adds an opt-in low balance email alert, sent when your credit balance drops below a threshold you set in USD (#54).
  • Settings -> Keys shows each key’s status and how much of its request and spend limits it has used (#55).
​

Product

  • Chat shows a banner with Create account and Sign in buttons when a signed-out visitor tries to send a message (#52).
​

API & SDKs

  • The AI SDK provider package is available on npm, starting at 1.0.1; see AI SDK (#48).
  • The TypeScript SDK 1.0.1 bundles devtools request capture, so installs no longer need a separate devtools core package (#47, #48).
​

Models

GLM 5 model release artwork
​
Last modified on October 28, 2026