TypingMind Release Notes

We have just added support for Kimi K3 through multiple providers: Moonshot, OpenRouter, Vercel AI Gateway.

Added a new Max reasoning effort option for supported AI models such as GPT-5.6 and some Claude models.

TypingMind now supports OpenAI’s new GPT-5.6 model family, giving users three options for different workloads:

• GPT-5.6 Sol — the flagship model, designed for complex reasoning, advanced coding, long-running agent workflows, scientific research, and cybersecurity tasks.
• GPT-5.6 Terra — a balanced, lower-cost model for everyday professional work, offering strong capability at approximately half Sol’s price.
• GPT-5.6 Luna — the fastest and most cost-efficient option, ideal for high-volume requests, quick answers, lightweight coding, summarization, and routine automation.

Claude Fable 5 is back!

Enable and use it on TypingMind now!

Bug Fixes and Improvements:

• Issue with Temporary Chat still being saved within folders: now Temporary Chat can no longer be turned on when the current chat belongs to a project/chat folder.
• Replaced DALL-E plugin access with GPT Image Editor for all plans.

Claude Sonnet 5 and Nano Banana 2 Lite are now live on TypingMind!

• Claude Sonnet 5: Anthropic’s most agentic Sonnet model yet.
• Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image): Google’s fastest and most cost-efficient image model.

New models have also been updated across providers such as OpenRouter and Vercel AI Gateway.

We’ve just added provider tool support for xAI models. You can now use Web Browser and Code Sandbox with Grok models.

In this update, we also added a temporary fix for the CORS issue affecting the Perplexity Search plugin and Perplexity models.

You can now use Sakana Fugu on TypingMind!

The **Fugu Ultra **model is said to match the performance of Fable and Mythos.

Try now!

Fixed

• Disable drag-and-drop uploads on Admin Panel when Upload Documents option is disabled. Now, uploaded documents are also counted toward usage limits.
• Fixed image uploads for AWS Bedrock Claude models by properly resolving URL-based images and showing clearer re-upload errors.
• Fixed PDF/image issues for Fireworks and Novita by disabling unsupported PDF support.

Improved

• Refreshed model catalogs and metadata across Chutes, DeepInfra, Fireworks, Mistral, Moonshot, Novita, OpenRouter, and Vercel.
• Improved OpenRouter API key verification and model search/import messaging.
• Improved MCP server display and connection reliability, including device-specific server filtering, better initialization, SSE timeout handling, and OAuth URL generation.

Added / Deprecated

• Added/promoted new models including GLM 5.2, Kimi K2.7 Code, MiniMax-M3, Qwen 3.7 Plus, and Fireworks fast-router variants and more.
• Deprecated older/free models.

We’ve refreshed the Add/Edit Custom Model form with a cleaner layout!

• Added automatic model capability checks before saving a custom model. TypingMind can now test model connectivity and detect support for features such as system role, streaming, tools, vision, thinking, prompt caching, supported parameters, and provider tools.
• Added support for storing cache read/write pricing fields for custom models.

You can now use Claude Fable 5 on TypingMind!

We have just added support for more AI providers, including Fireworks AI, Chutes AI, Novita AI, Minimax, Vercel Gateway and Xiaomi.

You can now connect your API keys and use models from these providers directly in TypingMind.

Go to Models → API Keys to set up your preferred provider and start using the available models.

Agent Skills are now supported on TypingMind!

TypingMind now supports Agent Skills — bundled procedural knowledge that helps AI perform specific tasks more effectively, such as SEO audits, copywriting, code reviews, brainstorming, and more.

Each Skill is packaged around a SKILL.md file, which contains task-specific instructions and optional reference files, following the open Agent Skills standard.

How to install Skills on TypingMind:

1. Go to Plugins.
2. Under the Store section, select Skills.
3. Install available Skills from the Store, or import your own Skill using a GitHub URL or ZIP file.

How to use Skills on TypingMind:

Enable the Skills you want to use, then start chatting. TypingMind will use the selected Skills to provide more specialized and task-focused responses.

Learn more about Skills

https://docs.typingmind.com/skills

Note: Skills are now available on TypingMind.com (Personal version). Support for TypingMind Teams coming soon.

We have just added Claude Opus 4.8 on TypingMind!

✨

Also in this update:

• Added Claude Opus 4.8 and Opus 4.8 (Fast) via OpenRouter.
• Updated model provider pricing/metadata for several catalogs, including DeepSeek, Deepinfra, Groq, Mistral, and OpenRouter models.

✨ New

• Added new Deepinfra models and OpenRouter models.
• Deprecated/deleted models are now kept as deprecated so old chats can show model status more clearly.

Improvements

• Updated OpenRouter and Deepinfra model metadata to reflect latest limits, pricing, and supported capabilities, including context/output limits, pricing, PDF support, and prompt caching support.
• Updated OpenRouter Anthropic latest aliases to use Claude-specific settings for better reasoning effort and prompt caching support.
• Expanded OpenRouter reasoning effort values: none, low, medium, high, xhigh.
• Removed deprecated DALL-E 3 builtin plugin as this image model is deprecated by OpenAI. Please switch to GPT Image Editor for image generation.

Fixed

• Fixed Fireworks provider issue where previous user input or stale/duplicated assistant content could be echoed when using the OpenAI Responses API format.