diff --git a/_data/leaderboard.csv b/_data/leaderboard.csv index 8bf83695c..79b40fe49 100644 --- a/_data/leaderboard.csv +++ b/_data/leaderboard.csv @@ -6,4 +6,5 @@ gemini-1.5-pro-latest,57.1,45.9,diff-fenced,aider --model gemini/gemini-1.5-pro- claude-3-sonnet-20240229,54.9,43.6,whole,aider --sonnet,0.25.0,a5f8076,3/6/24 Deepseek Coder,54.5,47,whole,aider --model openai/deepseek-coder,0.30.1,c07f793,4/29/24 gpt-3.5-turbo-0125,49.6,39.8,whole,aider --35turbo,0.22.0,da14474,2/2/24 -Command-R+,29.3,22.6,whole,aider --model command-r-plus,0.28.0,a06c927,4/20/24 \ No newline at end of file +Command-R+,29.3,22.6,whole,aider --model command-r-plus,0.28.0,a06c927,4/20/24 +Llama3 70B,49.2,38.6,diff,aider --model groq/llama3-70b-8192,0.32.0,b5bb453,5/3/24 \ No newline at end of file diff --git a/aider/models.py b/aider/models.py index 4a9182f96..8f1e90057 100644 --- a/aider/models.py +++ b/aider/models.py @@ -233,6 +233,7 @@ class Model: self.edit_format = "diff" self.use_repo_map = True self.send_undo_reply = True + self.examples_as_sys_msg = True return # <-- if "gpt-4-turbo" in model or ("gpt-4-" in model and "-preview" in model): diff --git a/docs/leaderboard.md b/docs/leaderboard.md index 880ed3430..79b9eda5f 100644 --- a/docs/leaderboard.md +++ b/docs/leaderboard.md @@ -5,23 +5,18 @@ Model - Final score - First try score - Edit format + Percent correct Command - Version - Git commit - Date + Edit format {% for row in site.data.leaderboard %} {{ row.model }} - {{ row.second }} - {{ row.first }} + {{ row.second }}% + `{{ row.command }}` {{ row.format }} - {{ row.command }} {% endfor %} diff --git a/docs/llms.md b/docs/llms.md index c6a8d60f4..28dc55ceb 100644 --- a/docs/llms.md +++ b/docs/llms.md @@ -14,7 +14,7 @@ Aider works with a number of **free** API providers: - Google's [Gemini 1.5 Pro](#gemini) is the most capable free model to use with aider, with code editing capabilities similar to GPT-3.5. -- You can use [Llama 3 70B on Groq](#llama3) which is comparable to GPT-3.5 in code editing performance. +- You can use [Llama 3 70B on Groq](#groq) which is comparable to GPT-3.5 in code editing performance. - The [Deepseek Coder](#deepseek) model works well with aider, comparable to GPT-3.5. Deepseek.com currently offers 5M free tokens of API usage. - Cohere also offers free API access to their [Command-R+ model](#cohere), which works with aider as a *very basic* coding assistant.