Add Tool Calling Evaluation section with BFCLv4 results

Added Tool Calling Evaluation section to the model card with Berkeley Function-Calling Leaderboard v4 (BFCLv4) results comparing this quantized model against the base nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16 model. Includes Overall, Single Turn (Non-Live and Live), Multi-Turn, and Agentic accuracy with recovery percentages.

README.md

@@ -306,3 +306,57 @@ The model was evaluated across instruct, reasoning, and coding tasks. Scores are
     </tr>
   </tbody>
 </table>
+
+### Tool Calling Evaluation
+
+The model was evaluated on tool calling tasks using the [Berkeley Function-Calling Leaderboard v4 (BFCLv4)](https://gorilla.cs.berkeley.edu/leaderboard.html).
+[vLLM](https://docs.vllm.ai/en/stable/) was used as the serving backend for all evaluations.
+
+<table>
+  <thead>
+    <tr>
+      <th>Category</th>
+      <th>Benchmark</th>
+      <th>nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16</th>
+      <th>RedHatAI/NVIDIA-Nemotron-3-Ultra-550B-A55B-quantized.w4a16<br>(this model)</th>
+      <th>Recovery</th>
+    </tr>
+  </thead>
+  <tbody>
+    <tr>
+      <td><strong>Overall</strong></td>
+      <td><strong>BFCLv4 Overall Acc</strong></td>
+      <td><strong>55.44</strong></td>
+      <td><strong>53.95</strong></td>
+      <td><strong>97.31%</strong></td>
+    </tr>
+    <tr>
+      <td>Single Turn</td>
+      <td>Non-Live Acc</td>
+      <td>45.00</td>
+      <td>44.58</td>
+      <td>99.07%</td>
+    </tr>
+    <tr>
+      <td>Single Turn</td>
+      <td>Live Acc</td>
+      <td>71.65</td>
+      <td>71.87</td>
+      <td>100.31%</td>
+    </tr>
+    <tr>
+      <td>Multi-Turn</td>
+      <td>Multi-Turn Acc</td>
+      <td>42.12</td>
+      <td>42.12</td>
+      <td>100.00%</td>
+    </tr>
+    <tr>
+      <td>Agentic</td>
+      <td>Agentic Acc</td>
+      <td>57.64</td>
+      <td>54.24</td>
+      <td>94.10%</td>
+    </tr>
+  </tbody>
+</table>