Update files

leikareipa · Dec 5, 2024 · 7fea3b2 · 7fea3b2
1 parent 9f4727c
commit 7fea3b2
Show file tree

Hide file tree

Showing 4 changed files with 20 additions and 3 deletions.
diff --git a/blog/briefly-evaluating-qwq-preview/content.md b/blog/briefly-evaluating-qwq-preview/content.md
@@ -849,5 +849,4 @@ Both QwQ and Qwen 2.5 Coder added a game-over condition, though I'd say QwQ's is
 
 ## Conclusions
 
-QwQ Preview does a decent job, often rising above Qwen 2.5 Coder and sometimes even Claude 3.5 Sonnet; but on the flip side eating up quite a few more tokens. Interesting to see the full QwQ down the road.
-
+QwQ Preview does a decent job, often rising above Qwen 2.5 and sometimes even Claude 3.5 Sonnet; but on the flip side eating up quite a few more tokens. It appears to be a specialist model in that some tasks it does very well and others not too great. Interesting to see the full QwQ down the road.
diff --git a/blog/briefly-evaluating-qwq-preview/index.html b/blog/briefly-evaluating-qwq-preview/index.html
@@ -907,7 +907,7 @@
                     </dokki-iframe>
 <p>Both QwQ and Qwen 2.5 Coder added a game-over condition, though I'd say QwQ's is better. They both also implemented right-clicking to place flags, but since the UI framework only supports left-click, that can't be tested. Overall, I wouldn't call either version finished, but getting there.</p>
 </dokki-subtopic><dokki-subtopic title="Conclusions">
-<p>QwQ Preview does a decent job, often rising above Qwen 2.5 Coder and sometimes even Claude 3.5 Sonnet; but on the flip side eating up quite a few more tokens. Interesting to see the full QwQ down the road.</p>
+<p>QwQ Preview does a decent job, often rising above Qwen 2.5 and sometimes even Claude 3.5 Sonnet; but on the flip side eating up quite a few more tokens. It appears to be a specialist model in that some tasks it does very well and others not too great. Interesting to see the full QwQ down the road.</p>
 </dokki-subtopic></dokki-topic>
 
                     </dokki-topics>

diff --git a/blog/testing-open-and-closed-llms-some-more/content.md b/blog/testing-open-and-closed-llms-some-more/content.md
@@ -267,6 +267,15 @@
             <td class="s0"></td>
             <td class="s2"></td>
         </tr>
+        <tr>
+            <td><x-size>47G</x-size><b>athene-v2</b>:72b-q4_K_M</td>
+            <td>20%</td>
+            <td class="s1"></td>
+            <td class="s0"></td>
+            <td class="s0"></td>
+            <td class="s0"></td>
+            <td class="s1"></td>
+        </tr>
         <tr>
             <td><x-size>12G</x-size><b>qwen-2.5-coder</b>:32b-instruct-q2_K</td>
             <td>10%</td>

diff --git a/blog/testing-open-and-closed-llms-some-more/index.html b/blog/testing-open-and-closed-llms-some-more/index.html
@@ -350,6 +350,15 @@
                     <td class="s0"></td>
                     <td class="s2"></td>
                 </tr>
+                <tr>
+                    <td><x-size>47G</x-size><b>athene-v2</b>:72b-q4_K_M</td>
+                    <td>20%</td>
+                    <td class="s1"></td>
+                    <td class="s0"></td>
+                    <td class="s0"></td>
+                    <td class="s0"></td>
+                    <td class="s1"></td>
+                </tr>
                 <tr>
                     <td><x-size>12G</x-size><b>qwen-2.5-coder</b>:32b-instruct-q2_K</td>
                     <td>10%</td>
Original file line number	Diff line number	Diff line change
Expand Up		@@ -849,5 +849,4 @@ Both QwQ and Qwen 2.5 Coder added a game-over condition, though I'd say QwQ's is

		## Conclusions

		QwQ Preview does a decent job, often rising above Qwen 2.5 Coder and sometimes even Claude 3.5 Sonnet; but on the flip side eating up quite a few more tokens. Interesting to see the full QwQ down the road.

		QwQ Preview does a decent job, often rising above Qwen 2.5 and sometimes even Claude 3.5 Sonnet; but on the flip side eating up quite a few more tokens. It appears to be a specialist model in that some tasks it does very well and others not too great. Interesting to see the full QwQ down the road.