TypingMind Release Notes

We’ve improved the multi-model experience:

• Support stopping and regenerating messages in multi-model responses

• Support pasting files and large text as file attachments when replying in multi-model threads.

• Show statistics such as token usage and estimated cost for each model.

• Fork a chat directly from a multi-model response message.

• When finalizing a multi-model response, you can now click Edit finalize instruction to customize the prompt used for that specific finalize request before picking the model to generate it.

Note: The default finalize instruction can now be customized in Settings → Advanced Settings → Internal Prompts under a new "Finalize multi-model responses" section (with a "Reset to default" option), synced across your devices via cloud sync.

3 new AI models are now available on TypingMind:

• Claude Fable 5.1: Anthropic’s most advanced model for demanding reasoning, long-horizon coding, deep research, and complex knowledge work.

• Gemini 3.8 Flash: Google’s newest high-speed workhorse, built for powerful reasoning, coding, autonomous agents, and complex workflows—while keeping Flash-level speed and efficiency.

• Muse Spark 1.3: Meta’s latest Muse model, improved for agentic workflows and coding, with a stronger ability to handle long-running tasks and multiple workflows in one conversation.

Try them now on TypingMind!

Ox Alpha turns out to be Z.ai’s model — GLM-3.5 Flash! And it’s now officially available on TypingMind.

We also added Qwen3.8 Flash - latest model from Alibaba:

Bug Fixes and Improvements:

• Added native PDF capability detection with a new "Support PDF Input" toggle when manually importing custom models.
• Fixed an issue where synced images could not be used with the DeepSeek provider.
• Fixed an issue where attachments were not accessible in forked chats.

Admins now can add/enable skills for your team members:

• Add skills:

• Enable skills / control skill visibility to specific user groups:

• Assign skills to Agents:

• DeepSeek V4 Flash now supports vision via DeepSeek V4 Flash Vision Exp

• Ox Alpha (Free) is available via OpenRouter and OpenCode Go

Added GLM-5.3 across multiple providers on TypingMind.

Bug fixes and improvements:

• Fixed an issue where new multi-model chats would revert to a single model after switching chats.
• Added support for the Responses API type for OpenCode Go.
• Added support for external image URLs when using OpenAI Responses models.

We have just added Gemini 3.7 Flash on TypingMind:

It’s now easier to delete all cloud data within a specific data type, such as Files, Chats, Prompts, Agents, and more.

We have added Grok 4.6 and DeepSeek V4 Pro 0813 across all providers on TypingMind.

We have just added a new provider - Meta and their latest model - Muse Spark 1.2. Try now on TypingMind!

• Added Qwen 3.8 Max from multiple providers on TypingMind:

• Update DeepSeek V4 Flash 0731 - the latest version across providers:

App improvements:

• Shared chats now support displaying multiple responses.
• Your device will now stay awake while a response is being generated, preventing long responses from failing due to the screen turning off.

We have just added Kimi K3 via multiple AI providers such as OpenCode Go, Fireworks, Ollama Cloud.

Bug Fixes and Improvements:

Bug fixes:

• Fixed an issue where plugins couldn't properly access synced blobs.
• Fixed an issue where images attached to shared chat links appeared broken for people viewing the link without an account.
• Fixed an issue where generated images would disappear after the streaming response finished.

Improvements:

• Added support for using external PDF URLs with OpenRouter models, instead of requiring inline file data.
• Removed the redundant "TypingMind AI Gateway & Org ID" entry from Settings → Advanced Settings, since these settings have already moved to the Models tab.
• Uninstalling a plugin now also clears its saved settings and enabled/disabled state, so reinstalling it starts fresh instead of restoring stale values.

You can now use Gemini 3.6 Flash and Gemini 3.5 Flash Lite on TypingMind!

You can now use models via OpenCode Go and Ollama Cloud on TypingMind!