OpenInterpreter
recommendation for making 'gpt-4o-mini' the default model over '4-turbo'
Is your feature request related to a problem? Please describe.
I saw that 4-turbo is currently the default model, the issue that it's outdated and expensive, which may lead to a negative first experience for new users
Describe the solution you'd like
I would like to see gpt-4o-mini become the new default model upon install, it's a newer model that's both cheap yet effective. It would allow for a lot of natural experimentation and learning without becoming unexpectedly expensive.
Describe alternatives you've considered
No response
Additional context
No response
Where did you see 3.5 turbo?
It's currently 4 turbo - https://github.com/OpenInterpreter/open-interpreter/blob/main/interpreter/core/llm/llm.py#L42
You can set the model with interpreter --model gpt-4o-mini or any of the models supported on litellm https://docs.litellm.ai/docs/providers
ah I see you're correct, the reason I was confused is I associated 'turbo' with 3.5
still, when looking at https://platform.openai.com/docs/models:
GPT-4 Turbo and GPT-4 | The previous set of high-intelligence models
GPT-4o is cheaper and faster than GPT-4 Turbo.
is there a reason to use turbo instead of 4o or 4o-mini as default?
| MODEL | DESCRIPTION | CONTEXT WINDOW | MAX OUTPUT TOKENS | TRAINING DATA |
|---|---|---|---|---|
| gpt-4o | GPT-4o: Our high-intelligence flagship model for complex, multi-step tasks. GPT-4o is cheaper and faster than GPT-4 Turbo. Currently points to gpt-4o-2024-05-13 [1]. |
128,000 tokens | 4,096 tokens | Up to Oct 2023 |
| MODEL | DESCRIPTION | CONTEXT WINDOW | MAX OUTPUT TOKENS | TRAINING DATA |
|---|---|---|---|---|
| gpt-4-turbo | The latest GPT-4 Turbo model with vision capabilities. Vision requests can now use JSON mode and function calling. Currently points to gpt-4-turbo-2024-04-09. |
128,000 tokens | 4,096 tokens | Up to Dec 2023 |
@drhouse we've discussed it and might change it, but I've found 4-turbo to be more performant than 4-o. However, all users are able to easily set whichever model they prefer
@MikeBirdTech interesting, I'm always looking to gain an edge in coding, I'm curious how has 4-turbo been more performant than 4o? Is there a place I can see the discussion that was had over this?
well, I've always used 4o-mini with interpreter in --os mode, and today as an experiment I tried using 4o. I racked up $5 in API usage in no time flat while using it try to fix a python package install. And, 4-turbo is more expensive than 4o. I'm not sure how people afford to use it regularly.
Of course, there is a question about a model's quality while doing tasks, but I use interpreter mainly for either non-coding or incidental coding tasks. For AI coding, I use more dedicated tools like Aider or Claude Dev.
Overall, I still think it would make sense to make 4o-mini the default model, striking the best balance between speed and cost. In any case, the future of Open Interpreter is bright.
I just wasted $6 because I ran open-interpreter --conversations and forgot to set the model to 4o-mini. 🤦
Default should be 4o-mini because it's cheapest and smart enough. If users want to spend more money, that should be opt-in, not opt-out.
