Many of you are excited about OpenAI’s release of their newest model: GPT-4o. You can already use our OpenAI modules such as “Create a Chat Completion” and “Analyze Images (Vision)” to use this new model. (For ‘Analyze Images’ module, simply map the model and type GPT-4o).
We are currently working to update the OpenAI Assistant-related modules to the Assistant v2 model that is necessary for working with GPT-4o and an assistant. This is a top priority for us, and our development team is working to make it happen as soon as possible.
I will update this thread with any news about the rollout as soon as it becomes available.
Thank you very much for your patience and understanding.
As a temporary workaround, you could use an older assistant model or use the “Make an API Call” module with the custom API call to OpenAI API with the “OpenAI-Beta: assistants=v2” header.
A guide on how to use the “Make an API Call” modules.
Hi, I tried the temporary workaround with the Make API Call module, adding the additional header information as directed. It does not fix the issue. Has anybody else had success with GPT-4o model?
Hey Tony, I don’t know if it is of any help to you, but I managed to get one of my assistants working with 4o yesterday. I think I’ve attached a screenshot of how I added an extra header (not replied on the forum before) + note that for the url you still use /v1/.
By referring to the API documentation provided by OpenAI, I successfully built an Assistant API that supports GPT-4o
within Make by structuring a three-layer scenario. Compared to the previous version of the Assistant API, the response speed has significantly improved.
In the second layer scenario, the “Create Run” process is executed, and in the third layer scenario, the “Watch Run” process is carried out. In the third layer, a module that pauses for five seconds is included to ensure there is enough time for the Assistant API to generate a message. This setup continues to repeat the same process until the status is marked as complete.
Hi Scotty, Many thanks for the reply! I did try to add an extra Header (as per the docs), but that did not work for me. Strange, as I’m using /v1/assistants endpoint too! As it’s not essential to have GPT-4o, we’re going to wait for the updates from Make.
At this time, we don’t have a specific date to share. However, we expect the update to be available in the upcoming days. As mentioned earlier, I will make sure to update you as soon as I have any new information.
Thank you for your patience and your excitement about this update.
Had one more thought. In your OpenAI assistant page (https://platform.openai.com/assistants/asst_#) you can choose the API version at the bottom of each assistant. I had to change mine to latest from v1 as v1 doesn’t work with 4o. That is my only other suggestion! Screengrab attached. Good luck:)
I’m happy to let you know that our OpenAI modules are now compatible with GPT-4o!
You can find it readily available in the ‘Create a Completion’ module. If you’re using the ‘Message an Assistant’ module instead, please update the module in your scenarios.
Thank you for being so patient and enjoy exploring the new capabilities!
I am using the assistant function and rely on this for some processes. I still cannot use GPT-4o in the assistant module it seems – I still get this error message:
“The operation failed with an error. [400] The requested model ‘gpt-4o’ cannot be used with the Assistants API in v1. Follow the migration guide to upgrade to v2: https://platform.openai.com/docs/assistants/migration.”
As you write that your modules (plural) are now compatible – how can I use the assistants API in v2 or use Assistants with GPT-4o?
First and foremost, I would like to extend my gratitude for developing the module for the Assistant API compatible with GPT-4o on such a tight schedule. I am currently testing the module you created and have reached out to request a specific improvement.
While you were developing the new Assistant API module for GPT-4o, I successfully leveraged a custom API module to construct scenarios equivalent to the Assistant API module. (Please refer to the attached images for more details.)
During this development process, I discovered that if the “tool_choice” parameter in the Http request for message generation does not specify “file_search,” accurate responses based on uploaded file information are occasionally not generated.
In the newly developed Assistant API module, this “tool_choice” parameter cannot be specified. From my recent tests, I have confirmed that accurate responses were not generated. Would it be possible to update the module to include this feature?
I do not believe the post announcing the new modules was very clear above as it implies you need to now use Create a Completion vs Message an Assistant. What I found after testing is there is a new Message an Assistant module which no longer says deprecated but otherwise looks similar and does work with the 4o agent. Sadly, it does appear you need to go in and replace every existing module with the new module and recreate all of your information vs just ‘upgrade’ the ones in use.
Does that mean we just need to pay for ChatGPT plus and we can use it as a message assistant module in Make? I’m thinking of investing in ChatGPT plus if this is the case.
Thanks @samliew,
How can I see ChatGPT 4o on Make? I’ve seen other creators that successfully did that with their chat completion module but unsure if it needs other set up, or just the regular API keys + Organization ID is good enough?
A guide on how to use the “Make an API Call” modules.
