> ## Documentation Index
> Fetch the complete documentation index at: https://docs.orq.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Release 3.14

> Release 3.14 adds Google Gemini 3 Pro Preview with 1M+ context window and Anthropic structured outputs with JSON schema validation.

<Update label="Google Gemini 3 Pro Preview" description="v3.14.0">
  Added support for Google's latest Gemini 3 Pro Preview model with advanced reasoning capabilities and configurable thinking levels.

  **Key Features:**

  * **Advanced reasoning** with configurable thinking levels (low/high) for complex problem-solving tasks
  * **1M+ context window** (1,048,576 tokens) for processing large documents and extended conversations
  * **Vision capabilities** supporting both text and image inputs
  * **Tool calling support** with JSON schema response format validation

  The model is available through both Google Vertex AI and Google AI providers, with full support across Deployments, Experiments, and the AI Gateway.

  <Note>
    Explore Gemini 3 Pro Preview in the [AI Router](/docs/model-garden/overview) and leverage its advanced reasoning capabilities for your most complex use cases.
  </Note>
</Update>

<Update label="Anthropic Structured Outputs Support" description="v3.14.0">
  Claude models now support JSON schema validation through Anthropic's Structured Outputs beta API, ensuring your AI responses conform to predefined schemas with guaranteed type safety.

  **What's included:**

  * **JSON Schema Validation** for Claude Sonnet 4.5 and Claude Opus 4.1, enabling strict adherence to your defined output structure
  * **Type-safe responses** that match your schema requirements, reducing parsing errors and validation overhead
  * **Seamless integration** across Deployments, Experiments, and the AI Gateway using the `json_schema` response format type

      <Note>
        Learn more about [Claude Structured Outputs](https://docs.claude.com/en/docs/build-with-claude/structured-outputs) including JSON schema limitations and best practices.
      </Note>
</Update>

<Update label="OpenAI GPT-5.1 Models" description="v3.14.0">
  Added support for OpenAI's latest GPT-5.1 models to the **AI Router**.

  **Available Models**

  * **gpt-5.1** - The best model for coding and agentic tasks with configurable reasoning effort. Features a 400k context window and advanced capabilities for tool calling and structured outputs.
  * **gpt-5.1-chat-latest** - GPT-5.1 snapshot currently used in ChatGPT with 128k context window. Supports vision capabilities and streaming.

  Both models support advanced features including JSON mode, JSON Schema response formats, strict tool calling, and developer messages for enhanced control over model behavior.

  <Note>
    Explore these new models in the [AI Router](/docs/model-garden/overview) and use them in Deployments, Experiments, or via the AI Gateway.
  </Note>
</Update>

<Update label="Home Dashboard Revamp" description="v3.14.0">
  <img src="https://mintcdn.com/orqai/QJ8YihZ2va_RnNmX/images/Screenshot2025-10-30at17.44.35.png?fit=max&auto=format&n=QJ8YihZ2va_RnNmX&q=85&s=d4d6266cae313242eaa30e72479c90ed" alt="Screenshot2025 10 30at17 44 35 Pn" width="1914" height="1224" data-path="images/Screenshot2025-10-30at17.44.35.png" />

  We have rolled out a major upgrade to the home dashboard, giving it a completely refreshed look and a more actionable experience.

  **What’s new:**

  * **Quick actions** to get started right away. Create a deployment, run an evaluation, open the playground, or start an experiment directly from the dashboard. These workflows used to live in the sidebar, but we have brought them front and center to make your next step obvious and effortless.
  * A redesigned layout with central dashboards and overview tables for a **clear snapshot of your workspace performance.**
  * A **Recently Viewed** section so you can quickly jump back into your most recent experiments, deployments, or files as your workspace grows.
  * **Detailed model and deployment insights** that help you understand which models drive requests, cost, and latency.

      <Note>
        Explore the new Dashboard features in detail at [Dashboard](/docs/analytics/dashboards).
      </Note>
</Update>

<Update label="Trace Automations" description="v3.14.0">
  You can now create self-updating datasets based on your traces using Trace Automations. Instead of manually sorting through the traces, you can define rules that continuously collect the relevant ones for analysis and experimentation.

  **What’s new**

  * **Advanced filtering** across any trace attribute such as metadata, model, provider, cost, tokens, or project. Combine multiple filters and apply sampling to focus only on the traces that matter.
  * **Add to Dataset** action that lets you automatically collect traces that match your filters. These datasets stay up to date as new traces arrive, so you can run experiments on the latest production data at any time.

  Additional actions including **Apply Evaluator** and **Add to Annotation Queue** are coming soon to help you extend automation into evaluation and review workflows.

  <Note>
    Want to set up Trace Automations? See [Trace Automations](/docs/observability/trace-automation) for setup steps, filtering options, and dataset configuration.
  </Note>
</Update>

<Update label="AI Router Revamp" description="v3.14.0">
  <img src="https://mintcdn.com/orqai/QJ8YihZ2va_RnNmX/images/Screenshot2025-10-30at18.41.00.png?fit=max&auto=format&n=QJ8YihZ2va_RnNmX&q=85&s=5373f4cf6dedacf8ebaab85a2d865a5c" alt="Screenshot2025 10 30at18 41 00 Pn" width="1658" height="574" data-path="images/Screenshot2025-10-30at18.41.00.png" />

  The **AI Router** has been completely reimagined to provide deeper insights and faster access to the right models for your use cases, making it easier than ever to explore, compare, and integrate them directly into production via the AI Gateway or use them in a Deployment or Experiment.

  **Key Features**

  * **Multimodality visibility** clearly shows which models support text, image, or both for input and output.
  * **Intelligence and speed ratings** help you compare models at a glance based on performance and capability.
  * **Comprehensive token and pricing data** display input and output costs to make budgeting and optimization simple.
  * **Filter and sort** controls across provider, model type, and activity status make model discovery effortless.
  * **AI Router code snippets** are available directly from each model card, letting you copy and deploy instantly through the Orq AI Router.
      <Note>
        Learn more about all available models and their capabilities at [AI Router](/docs/model-garden/overview).
      </Note>
</Update>

<Update label="Mistral and Custom ElevenLabs Integration" description="v3.14.0">
  Mistral and ElevenLabs added as new providers with Bring Your Own Key (BYOK) support

  <Note>
    To learn more about Mistral and ElevenLabs BYOK integration, see

    [Integrations](/docs/integrations/providers/overview) for setup instructions and Gateway configuration details.
  </Note>
</Update>

<Update label="Private OpenAI-Compatible Model" description="v3.14.0">
  You can now connect and configure your own OpenAI-compatible models directly in Orq. This gives you full flexibility to use custom, self-hosted, or fine-tuned models through the same interface and workflows you already use.

  **Key Components**

  * Custom model configuration for any OpenAI-compatible endpoint, including chat, completion, embedding, or image models.
  * Bring your own API key and base URL to connect private or hosted models securely through the Orq AI Gateway.
  * Full parameter control including temperature, max tokens, input and output token pricing, and region.
  * Add detailed model descriptions for easy discovery and team collaboration.

  **Integrated Observability**

  When your private model runs through the AI Gateway, all requests and responses are automatically traced in Orq. This gives you complete observability into model behavior and performance, while also enabling:

  * Dataset creation from real production traces
  * A/B testing between different models or configurations
  * Evaluator workflows to continuously measure quality and outcomes

      <Note>
        For detailed setup instructions and configuration options, see [OpenAI-compatible models](/docs/integrations/providers/open-ai-like).
      </Note>

    Alongside OpenAI-compatible models, you can also connect Azure deployments, Vertex AI models, and LiteLLM instances, each with full setup details available in the documentation.
</Update>

<Update label="Experiment Comparison View Update" description="v3.14.0">
  We have enhanced the comparison experience in Experiments to make it easier to analyze model performance side by side, even when working with large inputs or complex variable replacements.

  **Key Improvements**

  * Redesigned layout that moves the input variables and expected output to the left, giving you a cleaner overview when comparing multiple models or longer text outputs.
  * Evaluator section has been moved to the bottom of the comparison view, keeping the focus on model responses while maintaining quick access to evaluation scores and feedback.
  * Optimized for clarity and readability, especially when working with large datasets or extended completions.

  This update makes it simpler to review, compare, and evaluate your experiments at scale, ensuring that insights are easy to spot and interpret across models.

  <Note>
    Learn how to use the comparison view in [Running an Experiment](/docs/experiments/creating#running-an-experiment).
  </Note>
</Update>

<Update label="Structured Outputs" description="v3.14.0">
  You can now save and reuse JSON Schemas directly from your directory, ensuring consistent structured outputs across Deployments, Experiments, and Prompts.

  * Improved JSON editors and clearer validation

      <Note>
        To learn more about Structured Outputs, see [Creating a Deployment](/docs/deployments/creating)
      </Note>
</Update>

<Update label="UI and UX Improvements" description="v3.14.0">
  Dozens of refinements across the platform for a smoother and more consistent experience.

  * Enhanced error handling across prompts, tools, and experiments
  * Updated icons, spacing, and alignment
  * More responsive layouts across Playgrounds, Deployments, and Evaluators
</Update>

<Update label="Custom Reports" description="v3.14.0">
  <img src="https://mintcdn.com/orqai/QJ8YihZ2va_RnNmX/images/Screenshot2025-10-30at18.48.59.png?fit=max&auto=format&n=QJ8YihZ2va_RnNmX&q=85&s=4797010397bdf26fbafead26167aaa25" alt="Screenshot2025 10 30at18 48 59 Pn" width="1604" height="1396" data-path="images/Screenshot2025-10-30at18.48.59.png" />

  We have made a major update to the Reports section, introducing a flexible reporting system that provides a unified view of your system’s health. With off-the-shelf templates and advanced filtering, you can easily explore performance, cost, and reliability across projects, quickly identify bottlenecks or inefficiencies, and generate reports that align with your team’s KPIs for consistent, shareable insights across workspaces.

  **Key Capabilities:**

  * Customizable views with filters for time range and project selection, allowing you to focus on the data that matters most.
  * Build your own custom report by selecting and combining charts from different clusters.
  * Comprehensive insights with prebuilt charts for:
  * Cost breakdowns and model-provider splits
  * Error rates and anomaly detection
  * Response time by model and latency analysis
  * Request and token volume tracking

      <Note>
        Get started with Custom Reports at [Custom Reports](/docs/analytics/custom-reports).
      </Note>
</Update>