Skip to main content

What's New in Version 0.669.0

Released on February 3, 2026

This release introduces the highly anticipated Kimi K2.5 Thinking model, enhances voice agent controls, delivers significant upgrades to Sorcerer and Mystica families, improves code interpreter stability for enterprise databases, and expands MCP/API capabilities for intelligent agent optimization! 🚀

🧠 New Model: Kimi K2.5 Thinking​

Instant Hit with Record-Breaking Demand​

We're thrilled to introduce Kimi K2.5 Thinking, the latest addition to our model lineup! This model has taken the platform by storm, receiving an insane number of requests in less than 48 hours since its release. The response from our user community has been overwhelming, making this one of our fastest-adopted models ever.

Why the excitement?

  • Advanced Reasoning - Enhanced thinking capabilities for complex problem-solving
  • Superior Performance - Exceptional results across a wide range of tasks
  • Optimized Efficiency - Balanced performance and cost for production workloads

Kimi K2.5 Thinking is now available across all subscription tiers and ready to power your most demanding AI workflows.

Strategic Model Deprecation Initiative​

Alongside the introduction of Kimi K2.5 Thinking, we're announcing a strategic model deprecation program. We've identified and marked as deprecated a series of models that receive very low load and usage from our community.

Our Commitment to Compute Efficiency:

  • Aggressive Deprecation Strategy - We're taking a more proactive approach to retiring underutilized models
  • Compute Reallocation - Resources freed from low-usage models are redirected to support high-demand models like Kimi K2.5 Thinking
  • Better Performance - Ensuring optimal compute availability for the models our users actually use

This strategic decision allows us to maintain high performance and availability for popular models while streamlining our offering. Deprecated models will continue to function for a transition period, and users will receive advance notice before any model is fully retired.

🎙️ Voice Agents Interface Enhancements​

Pre-Conversation Input/Output Selection​

The voice agents interface now provides enhanced control over conversation methods. Users can now select both the input and output methods before the beginning of a conversation, preventing potential issues with voice interaction flow.

New Capabilities:

  • Input Method Selection - Choose your preferred input method (voice, text, or hybrid) before starting
  • Output Method Control - Select how the agent responds (voice, text, or both) upfront
  • Seamless Transitions - Pre-configured settings prevent mid-conversation mode switching issues
  • Better User Experience - Eliminates confusion and ensures smooth voice interactions from the first moment

This enhancement gives users full control over their voice interaction preferences, ensuring a more reliable and predictable experience when engaging with voice-enabled agents.

🧙 Agentic Capabilities Upgrade​

Sorcerer Family: Enhanced Planning & Behaviours​

The Sorcerer model family has received significant enhancements to its agentic capabilities, delivering smarter planning and more sophisticated agentic behaviours.

Key Improvements:

  • Advanced Planning Logic - Better multi-step reasoning and task decomposition
  • Improved Tool Orchestration - More intelligent selection and sequencing of tools
  • Adaptive Behaviour - Dynamic adjustment of strategies based on task context
  • Enhanced Context Awareness - Better understanding of complex requirements and constraints

Mystica Family: Superior Agentic Intelligence​

The Mystica model family has also been upgraded with substantial agentic enhancements, bringing more refined planning capabilities and intelligent behaviours.

What's New:

  • Refined Planning Algorithms - Optimized planning for complex, multi-stage workflows
  • Intelligent Decision Making - Better evaluation of options and execution paths
  • Enhanced Collaboration - Improved coordination when working with other agents and tools
  • Smarter Resource Utilization - More efficient use of available capabilities within reasoning threads

Both families now deliver more reliable, intelligent, and capable agentic experiences for your most complex automation needs.

💻 Code Interpreter Improvements​

Enhanced Database Stability​

The code interpreter has been significantly improved to provide better stability when interacting with enterprise databases programmatically.

Enterprise Database Enhancements:

  • SQL Server Stability - Improved connection handling and query execution reliability for Microsoft SQL Server
  • Oracle DB Support - Enhanced stability and performance when working with Oracle databases
  • Robust Error Handling - Better recovery from transient connection issues and database timeouts
  • Optimized Query Execution - More reliable execution of complex database procedures and transactions

These improvements ensure that agents using the code interpreter for database operations experience fewer interruptions and more consistent results, particularly in enterprise environments with complex database infrastructures.

🔌 MCP & API Expansions​

Intelligent Agent Optimization via Live Information​

We've made significant enhancements across the MCP (Model Context Protocol) server and API to enable agents to make more informed optimization decisions based on live platform information.

New Evaluation Capabilities:

  • Agent Blueprints Access - Agents can now retrieve and evaluate agent blueprints to understand optimal configurations
  • Available Models Intelligence - Real-time access to the models catalog with capabilities, performance characteristics, and availability status
  • Release Notes Integration - Agents can access and analyze the latest release notes to understand new features and optimizations
  • Intelligent Optimization - Based on live information and latest release notes, agents can now correctly evaluate how to optimize their own configurations and behaviours

Practical Applications:

  • Self-Optimizing Agents - Agents that can adjust their model selection based on current platform capabilities
  • Configuration Recommendations - Intelligent suggestions for blueprint improvements based on latest features
  • Feature Awareness - Agents that understand and leverage newly released capabilities automatically
  • Best Practice Enforcement - Automatic alignment with recommended configurations and patterns

This enhancement transforms agents from static configurations into intelligent, self-aware systems that continuously adapt to the evolving platform capabilities.

📋 Summary​

This update includes:

  • Kimi K2.5 Thinking - New model with record-breaking adoption in under 48 hours
  • Model Deprecation Program - Strategic retirement of low-usage models to optimize compute for popular models
  • Voice Agent Controls - Pre-conversation input/output method selection for smoother voice interactions
  • Sorcerer Enhancements - Significant upgrades to planning and agentic behaviours
  • Mystica Enhancements - Improved agentic intelligence and decision-making capabilities
  • Code Interpreter Stability - Better reliability with SQL Server and Oracle DB programmatic interactions
  • MCP/API Intelligence - Agents can now evaluate blueprints, models, and release notes for self-optimization

Thank you for using ToothFairy! We're constantly working to improve your experience. If you have any feedback or encounter any issues, please don't hesitate to reach out to our support team at support@toothfairyai.com.

Keep building magic with AI! 🧙‍♀️✨

v0.669.0