# 0.11.0

## Release Information

- OS support - Tested with Windows-x86 (10), Windows-arm64 (10), Mac-arm64 (14), Linux (Ubuntu 22.04)
- QAIRT SDK - Tested with versions (2.46.0, 2.47.0)
- Model Visualization format support - ONNX, TensorFlow, PyTorch (1.x, 2.x), TFLite, Executorch Program, DLC, EAIX, HTP, and QNN JSON

## Highlights

- Model diff - Graph View

    - The Diff panel now includes a visual graph representation of the diff between two models, complementing the existing diff table view
    - Added, removed, and modified nodes are visually indicated using color-coded highlights directly in the graph
    - Clicking a node in the diff graph navigates the diff table to the corresponding entry, and clicking a row in the diff table jumps the graph view to that element
- Model diff - Diff Stats

    - The Diff stats tab now displays aggregated diff results organized by operation and tensor, showing how many ops were added, removed, modified, or unchanged between the reference and comparison models
    - Results in the Diff stats table can be searched and sorted by operation type
- Graph Query - Report-Based Rules

    - Extends Graph Query with the ability to create rules that query for ops or tensors by their performance or accuracy metrics
- Graph Query - Import / Export Rules

    - Users can now import and export Graph Query rules as JSON files, allowing them to save, share, and reuse complex query configurations
- Analytics Overlay

    - Users can overlay a loaded performance report or accuracy report onto the graph as a heatmap, coloring nodes and edges according to where their metric values fall within a configurable data range
    - For performance reports, users select a **Stat** (min, max, or avg) to determine which data series is overlaid
    - For accuracy reports, users select an **Accuracy mode** and **Accuracy Comparator** to control which data series is visualized
    - Data range can be set to **Automatic** (derived from the report’s min and max values) or **Custom** (user-defined min and max)
    - A color gradient is applied to interpolate node colors across the data range, enabling fast visual identification of performance or accuracy hotspots directly in the graph
- Python 3.12 Op Tensor Mapping Support for Linux

    - Linux Python 3.12 support for Op Tensor Mapping feature has been added
    - Windows Python 3.12 support for Op Tensor Mapping will be added in Visualizer version 1.0.0

## Bug Fixes

- Fixed an issue where clicking on a bar in the accuracy report caused all panel views in the aggregate workspace to be updated, rather than only scrolling the graph panel to the corresponding tensor
- Fixed a model diff issue where tensors with many encodings were causing a memory violation, not allowing the model diff to render. Encodings are truncated to only show the first 10 values, however, comparisons are still made on the full encoding list

## Known Issues

- Diffing larger models in the QAIRT Visualizer Graph Panel (i.e., with 40K nodes or more or larger than 1.5 GB) may result in performance degradation or may fail to load altogether due to memory constraints. Upcoming releases will address these limitations to allow extra-large models to be loaded.

## Notices

- QAIRT Visualizer content may include China design and development contributions

Last Published: Jun 03, 2026

[Previous Topic
Notices](https://docs.qualcomm.com/bundle/publicresource/80-87189-1/topics/0_10_1.md)