Skip to main content

Gemini 4 Neon and Qwen 4 Max, a shorter model picker, and Babbily rebuilt for your phone

Gemini 4 Neon and Qwen 4 Max join the model picker, and the six Sunsetting models leave it.

Written by Babbily Support

Gemini 4 Neon and Qwen 4 Max join the model picker, and the six Sunsetting models leave it. On a phone, the composer now stays above the keyboard, the sidebar fills the screen, and Settings, menus and pages are simpler. Temporary Chat is easier to recognize and to leave, on every screen.

Gemini 4 Neon and Qwen 4 Max

  • Gemini 4 Neon joins the picker directly above Gemini 4 Carbon. It's Google's fast multimodal reasoning model for coding, agentic workflows and complex tasks. Choose Low, Medium or High reasoning. It has no fully-off setting, so with reasoning off it runs at Low. It reads images, PDFs, audio and video directly, has a context window of about 1 million tokens, and uses more of your API usage budget per token than Gemini 4 Carbon.

  • Qwen 4 Max takes Qwen 3.8 Max's place, last in the picker. It's Alibaba's flagship model for complex coding, knowledge work and agentic tasks, with a Reasoning toggle and a context window of about 1 million tokens. It reads images directly, and Babbily reads PDFs and other files for it.

The Sunsetting models leave the picker

GPT-6.1 Sol, GPT-6 Astra, Claude Fable 5.1, Gemini 3.8 Flash, Kimi K3 and Qwen 3.8 Max are no longer in the model picker, which now lists 19 models. Choose GPT-6.2 Sol, GPT-6.2 Astra, Claude Fable 5.2, Gemini 4 Argon, Kimi K3.1 Pro or Qwen 4 Max instead.

  • Chats that used one continue on another model. When you open one, Babbily selects Auto or another current model, and tells you if the model was your own choice. Older replies still show the model that wrote them.

  • Nothing is resent. If your last message didn't get a reply, choose Retry to send it with the new selection.

  • Scheduled Tasks that used one run on a replacement and say so. Open the task to choose a model yourself.

Auto uses Gemini 4 Neon for things you use

Requests for something to use rather than read, such as a calculator, a quiz, a checklist, a picker or a simulator, now go to Gemini 4 Neon at Medium reasoning. In our tests it built the more demanding ones better than the model it replaces, and matched it on everyday ones. Each of these requests uses about 2.5 to 4 times as much of your API usage budget as before. Simple requests still start on Claude Haiku 5.5, and when your remaining budget runs low Auto still favors models that use less. As before, Auto picks differently for a message with a file, in a project, with a selected skill or with a Canvas artifact open. See how Auto handles files, projects and other context.

Babbily on your phone

  • The composer stays above the keyboard. On iPhone and Android, the composer rides on top of the on-screen keyboard instead of hiding under it, in the same place in a new chat and in a thread. When you're at the end of a chat, the last reply stays in view as the keyboard opens or your draft grows.

  • A full-screen sidebar. Choose Open menu: the sidebar fills the screen when you hold your phone upright, and opens as a full-height panel when it's sideways. Search sits above New chat, News, Finance and Translate fold under More, and every chat in Recents shows its … menu.

  • Simpler Settings. Sections sit flat on the page, and Appearance shows Color mode and Chat font as rows.

  • A two-level + menu, as on a computer. Select + for your attachments, then Tool settings and Plugins, each with its own page and a Back button.

  • Clearer sheets. Search opens ready to type, Share ends on a Copy link button, the language menu shows each language in its own name above yours, and Log out sits apart in the account menu.

  • Artifacts and Projects. The Artifacts page shows its starters in one row you can scroll sideways, with its types behind one filter button, and Projects are a simple list.

  • Quieter surfaces. Menus, dialogs and sheets stand out by shade instead of shadow, and the browser's toolbar matches your color mode.

Temporary Chat

  • Leave from the header. While Temporary Chat is on, select the Temporary Chat button at the top right, then Exit and delete. Inside a chat on a phone, the button shows only its icons.

  • It's clear you're in one. A new temporary chat greets you with You're in a temporary chat, and the composer has a dashed border while the mode is on.

  • A dark card on your phone. Held upright, a temporary chat sits on a dark card under the header, in light mode too, and the menus that open from it are dark as well.

Voice Mode on a newer voice model

Voice Mode now runs on OpenAI's GPT-Live 1.1, which OpenAI says understands speech more accurately and keeps a steadier accent across languages. A call that started before this update keeps the earlier model until it ends.

  • Code goes to your chat model. Asking for code in a call, even a short function, now goes to your chat model instead of being read aloud.

  • So do your memory and drafts. Questions about your memory and earlier chats, asking Babbily to remember something, and drafting an email, a message or a post still go to your chat model, as lookups do.

  • Short answers stay quick. Greetings, quick questions, jokes, and questions about what was already said in the call or the chat are still answered by voice directly.

  • Same voices, same cost. The 31 voices are unchanged, and a call uses your API usage budget at the same rate as before.

Voice Mode is available on paid plans. See what a call hands to your chat model.

Also in this release

  • Quieter replies everywhere. The action icons under your messages and Babbily's replies stay muted until you point to one, and code blocks have no border.

Did this answer your question?