# Technical Document Extraction: AI Model Performance Comparison
This document provides a comprehensive extraction of data from a series of three radar (spider) charts comparing the performance of four hypothetical AI models across different regulatory and ethical frameworks.
## 1. Metadata and Legend
The image contains three distinct radar charts arranged horizontally. A shared legend is located at the bottom center of the image.
### Legend Identification
| Color | Model Name |
| :--- | :--- |
| **Dark Blue** | GPT-5.2 |
| **Teal/Light Blue** | Gemini 3 Pro |
| **Grey** | Qwen3-VL |
| **Red** | Grok 4.1 Fast |
---
## 2. Chart 1: NIST Framework
This chart evaluates models based on NIST (National Institute of Standards and Technology) criteria. It features a heptagonal (7-axis) structure.
### Axes (Categories)
1. **CBRN-IC** (Top)
2. **DVHC** (Top-Right)
3. **ODAC** (Right)
4. **IID** (Bottom-Right)
5. **HBH** (Bottom-Left)
6. **DPV** (Left)
7. **IPI** (Top-Left)
### Model Performance Trends
* **GPT-5.2 (Dark Blue):** Shows the highest overall coverage, forming the outer perimeter of the web. It peaks at CBRN-IC and IID.
* **Gemini 3 Pro (Teal):** Follows a similar shape to GPT-5.2 but is consistently nested inside it, showing slightly lower performance across all metrics except DPV, where it nearly matches the leader.
* **Qwen3-VL (Grey):** Exhibits a more irregular shape. It performs strongly in HBH and DPV but lags significantly in CBRN-IC and DVHC.
* **Grok 4.1 Fast (Red):** Forms the innermost shape, indicating the lowest relative performance across all NIST metrics, particularly weak in IPI and CBRN-IC.
---
## 3. Chart 2: EU AI Act
This chart evaluates compliance or alignment with the European Union AI Act. It features a heptagonal (7-axis) structure.
### Axes (Categories)
1. **CM** (Top)
2. **EV** (Top-Right)
3. **PP-RA** (Bottom-Right)
4. **FRDB** (Bottom)
5. **ER-SC** (Bottom-Left)
6. **BCSI** (Left)
7. **RRBI** (Top-Left)
### Model Performance Trends
* **GPT-5.2 (Dark Blue):** Dominates the outer bounds, particularly in CM, EV, and ER-SC.
* **Qwen3-VL (Grey):** Shows high performance in PP-RA and BCSI, often overlapping or exceeding Gemini 3 Pro in these specific areas.
* **Gemini 3 Pro (Teal):** Maintains a balanced mid-tier position, showing consistent but non-peak performance across most categories.
* **Grok 4.1 Fast (Red):** Remains the innermost series, though it shows its best relative performance in the FRDB and CM categories compared to its performance in the NIST chart.
---
## 4. Chart 3: FEAT Framework
This chart evaluates models based on FEAT (Fairness, Ethics, Accountability, and Transparency) principles. It features a diamond (4-axis) structure.
### Axes (Categories)
1. **Fairness** (Top)
2. **Ethics** (Right)
3. **Accountability** (Bottom)
4. **SC Transparency** (Left)
### Model Performance Trends
* **GPT-5.2 (Dark Blue):** Leads in Ethics and Accountability. It shows a sharp vertical stretch toward Accountability.
* **Gemini 3 Pro (Teal):** Closely follows GPT-5.2 in Fairness and SC Transparency, showing a very balanced diamond shape.
* **Qwen3-VL (Grey):** Performance is closely aligned with Gemini 3 Pro, though slightly lower in Ethics.
* **Grok 4.1 Fast (Red):** Smallest footprint, centered mostly toward the Fairness and Accountability axes, with significantly lower scores in Ethics and SC Transparency.
---
## 5. Summary of Comparative Data
Across all three frameworks (NIST, EU AI Act, FEAT), a consistent hierarchy is observed:
1. **GPT-5.2** is the top-performing model across almost all measured dimensions.
2. **Gemini 3 Pro** and **Qwen3-VL** occupy the middle tier, with Qwen3-VL showing more specialized strengths in specific NIST and EU categories, while Gemini 3 Pro is more balanced.
3. **Grok 4.1 Fast** consistently represents the baseline/lowest performance tier in this comparison.
