Updated benchmarks format

Author: cybermaggedon

benchmarks.md

@@ -22,15 +22,95 @@ measuring:
 
 ### Results
 
-| Platform | GPU | Model | Flow | Source material | In tok/s | Out tok/s | Total tok/s |
-| -------- | --- | ----- | ---- | --------------- | ---------- | ----------- | ---------- |
-| vLLM on Intel Gaudi 2 | Gaudi 2, 8 cards | meta-llama/Llama-3.3-70B-Instruct | document-rag+graph-rag | NASA Challenger Report Volume 1 | 1493.6 | 1545.8 | 3039.5 |
-| vllm-server | H100-SXM5-80GB (Tensordock) | TheBloke/Mistral-7B-v0.1-AWQ | document-rag+graph-rag | NASA Challenger Report Volume 1 | 304.3 | 1845.6 | 2150.0 |
-| VertexAI | n/a | Gemini 2.0 Flash | document-rag+graph-rag | NASA Challenger Report Volume 1 | 216.2 | 155.8 | 372.0 |
-| LMStudio | Radeon RX 7900 XTX | Gemma3 4B QAT | document-rag+graph-rag | NASA Challenger Report Volume 1 | 116.2 | 133.9 | 250.1 |
-| Granite Ridge | 128 Xeon Gen 6 CPU | mistralai/Mistral-7B-Instruct-v0.3 | document-rag+graph-rag | NASA Challenger Report Volume 1 | 117.7 | 90.0 | 207.8 |
-| LMStudio | Radeon RX 7900 XTX | Gemma2 9B | document-rag+graph-rag | NASA Challenger Report Volume 1 | 119.6 | 73.0 | 192.6 |
-| Granite Ridge | 128 Xeon Gen 6 CPU | meta-llama/Llama-3.3-70B-Instruct | document-rag+graph-rag | NASA Challenger Report Volume 1 | 67.0 | 22.4 | 89.3 |
+<table>
+  <thead>
+    <tr>
+      <th>Platform</th>
+      <th>GPU</th>
+      <th>Model</th>
+      <th>Config</th>
+      <th>Token Rate</th>
+      <th>Time to Process</th>
+    </tr>
+  </thead>
+  <tbody>
+    <tr>
+      <td>vLLM on Intel Gaudi 2 🏆</td>
+      <td>Gaudi 2, 8 cards</td>
+      <td>meta-llama/Llama-3.3-70B-Instruct</td>
+      <td>TC1</td>
+      <td>In: 1493.6<br/>Out: 1545.8<br/>Total: 3039.5</td>
+      <td>8.5 min</td>
+    </tr>
+    <tr>
+      <td>vllm-server on NVidia</td>
+      <td>H100-SXM5-80GB (Tensordock)</td>
+      <td>TheBloke/Mistral-7B-v0.1-AWQ</td>
+      <td>TC1</td>
+      <td>In: 304.3<br/>Out: 1845.6<br/>Total: 2150.0</td>
+      <td>12.0 min</td>
+    </tr>
+    <tr>
+      <td>VertexAI</td>
+      <td>n/a</td>
+      <td>Gemini 2.0 Flash</td>
+      <td>TC1</td>
+      <td>In: 216.2<br/>Out: 155.8<br/>Total: 372.0</td>
+      <td>69.4 min</td>
+    </tr>
+    <tr>
+      <td>LMStudio</td>
+      <td>Radeon RX 7900 XTX</td>
+      <td>Gemma3 4B QAT</td>
+      <td>TC1</td>
+      <td>In: 116.2<br/>Out: 133.9<br/>Total: 250.1</td>
+      <td>103.3 min</td>
+    </tr>
+    <tr>
+      <td>Granite Ridge</td>
+      <td>128 Xeon Gen 6 CPU</td>
+      <td>mistralai/Mistral-7B-Instruct-v0.3</td>
+      <td>TC1</td>
+      <td>In: 117.7<br/>Out: 90.0<br/>Total: 207.8</td>
+      <td>124.3 min</td>
+    </tr>
+    <tr>
+      <td>LMStudio</td>
+      <td>Radeon RX 7900 XTX</td>
+      <td>Gemma2 9B</td>
+      <td>TC1</td>
+      <td>In: 119.6<br/>Out: 73.0<br/>Total: 192.6</td>
+      <td>134.1 min</td>
+    </tr>
+    <tr>
+      <td>Granite Ridge</td>
+      <td>128 Xeon Gen 6 CPU</td>
+      <td>meta-llama/Llama-3.3-70B-Instruct</td>
+      <td>TC1</td>
+      <td>In: 67.0<br/>Out: 22.4<br/>Total: 89.3</td>
+      <td>289.3 min</td>
+    </tr>
+  </tbody>
+</table>
+
+### Test Configurations
+
+<table>
+  <thead>
+    <tr>
+      <th>Config ID</th>
+      <th>Flow</th>
+      <th>Source Material</th>
+    </tr>
+  </thead>
+  <tbody>
+    <tr>
+      <td>TC1</td>
+      <td>document-rag+graph-rag</td>
+      <td>NASA Challenger Report Volume 1 (1,549,890 tokens)</td>
+    </tr>
+  </tbody>
+</table>
 
 ## Procedure