From ee5d72301ae31ceb435ae984618f8ce0cf8069cc Mon Sep 17 00:00:00 2001
From: Paul Gauthier <paul@aider.chat>
Date: Fri, 24 Jan 2025 08:47:25 -0800
Subject: [PATCH] copy

---
 aider/models.py                              |  2 +-
 aider/website/_data/r1_architect.yml         | 10 +++---
 aider/website/_posts/2025-01-24-r1-sonnet.md | 34 ++++++++++++++++++++
 3 files changed, 40 insertions(+), 6 deletions(-)

diff --git a/aider/models.py b/aider/models.py
index 6e91d6aab..97575993b 100644
--- a/aider/models.py
+++ b/aider/models.py
@@ -841,7 +841,7 @@ MODEL_SETTINGS = [
         use_repo_map=True,
         streaming=False,
         use_temperature=False,
-        extra_params=dict(extra_body=dict(reasoning_effort="high")),
+        # extra_params=dict(extra_body=dict(reasoning_effort="high")),
     ),
     ModelSettings(
         "openrouter/qwen/qwen-2.5-coder-32b-instruct",
diff --git a/aider/website/_data/r1_architect.yml b/aider/website/_data/r1_architect.yml
index c036c7de8..5467e0fc7 100644
--- a/aider/website/_data/r1_architect.yml
+++ b/aider/website/_data/r1_architect.yml
@@ -23,7 +23,7 @@
   exhausted_context_windows: 0
   test_timeouts: 5
   total_tests: 225
-  command: aider --model deepseek/deepseek-reasoner
+  command: aider --architect --model r1 --editor-model sonnet
   date: 2025-01-23
   versions: 0.72.3.dev
   seconds_per_case: 251.6
@@ -49,7 +49,7 @@
   exhausted_context_windows: 1
   test_timeouts: 5
   total_tests: 225
-  command: aider --model deepseek/deepseek-reasoner
+  command: aider --model r1
   date: 2025-01-20
   versions: 0.71.2.dev
   seconds_per_case: 113.7
@@ -76,7 +76,7 @@
   exhausted_context_windows: 0
   test_timeouts: 2
   total_tests: 225
-  command: aider --model openrouter/openai/o1
+  command: aider --model o1
   date: 2024-12-21
   versions: 0.69.2.dev
   seconds_per_case: 133.2
@@ -103,7 +103,7 @@
   exhausted_context_windows: 0
   test_timeouts: 8
   total_tests: 225
-  command: aider --model deepseek/deepseek-chat
+  command: aider --model deepseek
   date: 2024-12-25
   versions: 0.69.2.dev
   seconds_per_case: 34.8
@@ -131,7 +131,7 @@
   exhausted_context_windows: 1
   test_timeouts: 8
   total_tests: 225
-  command: aider --model claude-3-5-sonnet-20241022
+  command: aider --model sonnet
   date: 2025-01-17
   versions: 0.71.2.dev
   seconds_per_case: 21.4
diff --git a/aider/website/_posts/2025-01-24-r1-sonnet.md b/aider/website/_posts/2025-01-24-r1-sonnet.md
index ac74ee7b5..c7d737e76 100644
--- a/aider/website/_posts/2025-01-24-r1-sonnet.md
+++ b/aider/website/_posts/2025-01-24-r1-sonnet.md
@@ -29,6 +29,40 @@ This is in contrast to the first wave of thinking models like o1-preview and o1-
 which improved when paired with many different editor models.
 
 
+## Try it
+
+Once you [install aider](https://aider.chat/docs/install.html),
+you can use aider, R1 and Sonnet like this:
+
+```bash
+export DEEPSEEK_API_KEY=<your-key>
+export ANTHROPIC_API_KEY=<your-key>
+
+aider --architect --model r1 --editor-model sonnet
+```
+
+Or if you have an [OpenRouter](https://openrouter.ai) account:
+
+```bash
+export OPENROUTER_API_KEY=<your-key>
+
+aider --architect --model openrouter/deepseek/deepseek-r1 --editor-model openrouter/anthropic/claude-3.5-sonnet
+```
+
+## Thinking output
+
+There has been 
+[some recent discussion](https://github.com/Aider-AI/aider/pull/2973)
+about extracting the `<think>` tokens from R1's responses
+and feeding them to Sonnet.
+That was an interesting experiment, for sure.
+
+To be clear, the results above are *not* using R1's thinking tokens, just the normal
+final output. 
+R1 is configured in aider's standard architect role with Sonnet as editor.
+The benchmark results that used the thinking tokens appear to be worse than
+the architect/editor results shared here.
+
 ## Results
 
 <table style="width: 100%; max-width: 800px; margin: auto; border-collapse: collapse; box-shadow: 0 2px 4px rgba(0,0,0,0.1); font-size: 14px;">