Open AI Workpace
Home Release notes Version 1.66

Version 1.66

23rd December 2024

This release introduces OpenAI's o1 model, Gemini 2.0 Flash Thinking, and a new way to make any model reason over hard problems with the /think skill.

We have also added minor performance improvements for workspace computer tasks.

What's new

  • o1: OpenAI's o1 model is faster than o1-preview and demonstrates enhanced mathematical and reasoning capabilities over o1-preview. In our initial testing we've found it has significantly improved at coding and creative writing.
  • /think skill: the new /think skill now allows any AI model to reason through complex problems methodically. This means you're not limited to just o1 for advanced reasoning - you can now use your preferred model and activate deep thinking capabilities when needed. Whether you're working on challenging mathematical problems, complex coding tasks, or need thorough analysis, the /think skill enables more deliberate and structured reasoning across all our supported models. When using the /think skill (including with o1) you can select between low, medium, or high reasoning effort to control the amount of tokens dedicated to reasoning. If you don't use the thinking skill when using a reasoning based model we automatically select medium.
  • Gemini 2.0 Flash Thinking: we're also adding support for Gemini 2.0 Flash Thinking, Google's experimental model, to your workspace. Similar to o1, this model focuses on transparent reasoning, showing its thought process while solving complex problems. While it's still in experimental mode, you can now try it out alongside our other models to see how it approaches different challenges. Like o1, it takes additional time to process responses in exchange for more detailed reasoning, giving you another option for tackling complex tasks.

What's next?

  • The holiday update:
    • Perplexity Search
    • Create with Code
    • Code focus
    • Kling Text to Video Creation
    • Mixmax Text to Video Creation
    • Create songs with Udio & Suno
    • Auto context length maximization
    • Various performance fixes
  • Workspaces: add a team or your family to a single workspace
  • AI agents: create task based or autonomous agents
  • Assistant detail screen to see sessions with a specific agent and make quick changes to an agents knowledge.
  • Threading to focus context window on a path
  • Improved RAG sources including the ability to see in a file where something was sourced from (text, audio, video)
  • Data sources: connect to Google Drive, Github, OneDrive + more.
  • Plus a lot more!