New models: GPT-5.6, Grok 4.5, and Muse Spark 1.1
Three new frontier model families are available now in the Babbily model picker.The coding results below are from CursorBench 3.2, Cursor’s evaluation of agents on ambiguous, multi-file tasks from real Cursor sessions. Scores and average cost per task are a July 9, 2026 snapshot and may change as the evaluation is updated.GPT-5.6 Sol, Terra, and Luna. GPT-5.6 Sol Max ranks third overall with a 67.2% score, behind Claude Fable 5 Max at 70.5% and Claude Fable 5 Extra High at 68.4%. GPT-5.6 Terra Max scores 64.9% at 5.22. Choose Luna when you want the faster, lower-cost GPT-5.6 option.Grok 4.5. Grok 4.5 High scores 66.7% and ranks fourth overall, half a percentage point behind GPT-5.6 Sol Max. Its average cost is 5.22 for Sol Max—about 3.5 times less for a similar score.Muse Spark 1.1. Use Meta’s upgraded model for coding, debugging, long-running agentic tasks, and multimodal reasoning across supported inputs.Compare model capabilities
Choose a model based on your task, speed, context, tools, and API usage budget.
Review CursorBench 3.2
See Cursor’s current scores, average per-task costs, methodology, and Grok 4.5 footnote.
Get notified when Babbily is done
Babbily can now send a push notification when a chat, image, or video finishes running in the background.What’s new
Turn notifications on when prompted. After your first prompt, Babbily may offer Get notified when Babbily is done. Choose Allow notifications, then allow notifications in your browser or device prompt.Manage notifications anytime. Open Settings > Account > Notifications to turn Chat completion notifications on or off.Step away while work runs. If you leave the tab or Babbily is running in the background, Babbily can let you know when the work is ready.Helpful for longer work. Completion notifications are especially useful for longer generations, research-heavy chats, image creation, and video runs.Documentation is easier to find. Settings now includes a direct Documentation link under Learn more.Notification settings
Turn background completion notifications on or off.
Browser and device support
Troubleshoot notification permissions and supported browsers.
Claude Sonnet 5 is now available in Babbily
Claude Sonnet 5 is now available in the Babbily model picker. It is the recommended Sonnet model for deep reasoning, writing, coding, analysis, and complex multi-step work.Sonnet 4.6 is still available, but it now appears as Sunsetting while teams move to Sonnet 5.Also added
GLM 5.2 Fast mode. GLM 5.2 now includes a Fast mode option when you want quicker responses for supported work.Nano Banana 2 Lite. Use Nano Banana 2 Lite when you want lighter-weight image generation and faster iteration.Updated model ordering. The model selector now places newer recommended models higher in the catalog.Model capabilities
Choose a model for reasoning, coding, speed, files, media, and API usage budget.
Image and video generation
Create images and videos with supported models and tools.
Sign in faster with passkeys
Babbily now supports passkeys for faster, passwordless sign-in on supported browsers and devices.You can add a passkey using Face ID, Touch ID, Windows Hello, a password manager, or a hardware security key. Once added, you’ll see Sign in with passkey on the web login screen.New users may be prompted to add a passkey during setup, and existing users can add, rename, or remove passkeys anytime from Settings > Account.Passkeys are currently available on web browsers. Native app passkey support will come later.Manage account access
Learn how to sign in, add passkeys, and manage account access.
News is faster, and Finance is now a dedicated workspace
Babbily now includes a full Finance area for market research, company pages, earnings, prediction markets, and watchlists. News also has richer discovery, source-backed article generation, local context, and market highlights in the side rail.What’s new
Finance in Babbily. Open Finance to review US markets, crypto, earnings, prediction markets, market summaries, company pages, price charts, key issues, analyst context, peers, and historical data where available.Watchlists and company research. Add companies to a watchlist, open company pages from market rows, and ask follow-up questions from a Finance page with the page context already attached.Richer News discovery. News now supports topic tabs, location-aware discovery, source citations, generated article views, article follow-up chat, weather, market outlook, and trending company links.Better performance and fallback behavior. News and Finance data now refresh more smoothly, keep recent results available during temporary provider issues, and load common market and news panels faster.News guide
Browse News, open sourced articles, update location, and continue from an article in chat.
Finance guide
Use market overviews, company pages, watchlists, earnings, prediction markets, and Finance research.
Sakana Fugu Ultra, Kimi K2.7, and GLM 5.2 are now available
Babbily now includes three more models for advanced reasoning, coding, long-context work, and agentic task execution.What’s new
Sakana Fugu Ultra. Use Fugu Ultra for complex prompts that benefit from multiple perspectives, deeper reasoning, and careful synthesis across approaches.Kimi K2.7. Use Kimi K2.7 for coding-focused work, debugging, multi-file reasoning, long-context development, and agentic software tasks.GLM 5.2. Use GLM 5.2 for long-horizon software engineering, repo-scale reasoning, refactoring support, and technical planning.These models are available now in Babbily.Read the full release note
See the public release note for Sakana Fugu Ultra, Kimi K2.7, and GLM 5.2.
New: Canvas Mode in Babbily 1.04
Babbily now includes Canvas, a dedicated workspace for creating, editing, and refining AI-generated work without leaving chat.What’s new
Edit generated work in Canvas. Drafts, tables, code, image briefs, and structured outputs can open in a durable side panel where you can keep working.Keep your edits in context. When you continue prompting in the same thread, Babbily can use the latest Canvas artifact state instead of only the original chat reply.Render structured outputs. Babbily can turn supported structured JSON outputs into cleaner visual components such as dashboards, tables, metrics, and callouts.Export or send work onward. Supported artifacts can be downloaded as Word, PDF, CSV, source, or JSON files. Some artifacts can also be sent to connected apps when connector actions are available.Read the full release note
See the public release note for Canvas Mode in Babbily 1.04.
Claude Fable added to Babbily
Claude Fable 5 is now available on Babbily — with a 1 million token context window.Fable is a Mythos-class model built for long-running, complex, and asynchronous work, with robust safeguards designed for serious tasks. That means fewer interruptions, less context loss, and more room to think through big projects from start to finish.Available now on Babbily.New features
Skills, tools, memory, auto mode, and connectors. Babbily can now personalize replies with persistent memory, use tools and skills automatically when they help, and pull live context from connected apps. Auto mode can choose available capabilities when they seem useful.API usage billing. A new monthly API usage budget replaces fixed message limits. Track what you’ve used, see what counts, and know when it resets — all from the chat sidebar.Updates
Clearer context indicator. The context window tracker now updates more reliably during long chats so you can tell when Babbily will compact older turns.Budget refresh. Your usage view refreshes faster after each turn so recent activity is reflected quickly; final usage may continue to settle as billing data refreshes.Fixes
- Resolved a stall when the context-limit warning appeared mid-send.
- Restored model fallback so chat keeps working when a preferred model is briefly unavailable.
- Fixed scrolling and layout glitches in temporary chats and on tablet viewports.
- Sign-in via Google and Apple is now more reliable on mobile.